bots-garden/mini-me · f0565c3

added .tickets/issues/0108-plan-store-shared-by-every-session.yaml +106 -0 Viewed

new file mode 100644
@@ -0,0 +1,106 @@
	1	+# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
	2	+id: 108
	3	+title: "plan: one store for the whole process — every ACP session sees the same plans"
	4	+state: closed
	5	+author:
	6	+ name: k33g
	7	+ email: ph.charriere@gmail.com
	8	+createdAt: 2026-10-21T08:00:00.000Z
	9	+updatedAt: 2026-10-21T10:00:00.000Z
	10	+labels:
	11	+ - kind::bug
	12	+milestone: v0.9.0
	13	+epic: Builtin tools
	14	+related:
	15	+ - 109
	16	+ - 110
	17	+body: \|
	18	+ ## The problem
	19	+
	20	+ The `plan` tool keeps its plans in memory, in ONE store for the whole mm
	21	+ process: `internal/tools/plan.go`, `getSessionID()` returns the constant
	22	+ "current-session" ("v1: use a single session for simplicity"). Every plan
	23	+ lives under that one key.
	24	+
	25	+ In the line mode and the window this goes unnoticed: one process, one
	26	+ conversation at a time. Over ACP it does not hold: one `mm -acp` process
	27	+ serves every thread the editor opens (Zed: each thread is a session/new),
	28	+ and they all share that one store. Concretely, with two threads open in
	29	+ Zed:
	30	+
	31	+ - thread A asks "do these 3 things…": the model creates `user-request-1`;
	32	+ - thread B, about something else, asks for a plan too: its unnamed plan
	33	+ becomes `user-request-2` (the name counts A's plan), and
	34	+ `plan list` in B answers with BOTH plans;
	35	+ - the model in B can read, `update`, `next` or `delete` A's plan — a small
	36	+ model told "Found 2 plan(s)" will happily work on the wrong one;
	37	+ - a `create` with a name A already used fails with
	38	+ `plan 'x' already exists`, for a thread that never created it.
	39	+
	40	+ Nothing tells the user or the model that the plans come from another
	41	+ conversation. It is also why plans cannot be written to the session file
	42	+ (the code's own "v2" note): they are not attached to a session at all.
	43	+
	44	+ ## What fixing it solves
	45	+
	46	+ - Each conversation has its own plans: one Zed thread can no longer see,
	47	+ change or delete another thread's plan, and plan names restart at
	48	+ `user-request-1` in each.
	49	+ - A prerequisite for the two follow-ups: /new clearing the right plans in
	50	+ every front end (ticket 109), and later keeping plans in the session file
	51	+ so /resume and session/load bring them back.
	52	+
	53	+ ## Direction (to confirm)
	54	+
	55	+ Key the store by the conversation the call belongs to. The tool does not
	56	+ know it today: the session id is not passed to tools (the same gap
	57	+ search_memory worked around in ticket 48). Options: a session id in the
	58	+ tool's context, set by each front end next to ui.SetActive; or one plan
	59	+ store per front-end session, handed to the tool through the active sink.
	60	+tasks:
	61	+ - id: 1
	62	+ title: "Choose how a tool call learns its conversation (context value vs the active sink) — the same gap as ticket 48's search_memory"
	63	+ state: closed
	64	+ priority: none
	65	+ author:
	66	+ name: k33g
	67	+ email: ph.charriere@gmail.com
	68	+ createdAt: 2026-10-21T08:00:00.000Z
	69	+ updatedAt: 2026-10-21T08:00:00.000Z
	70	+ - id: 2
	71	+ title: "Key the plan store by that conversation; plans of one ACP session invisible to another — with a test of two sessions over the wire"
	72	+ state: closed
	73	+ priority: none
	74	+ author:
	75	+ name: k33g
	76	+ email: ph.charriere@gmail.com
	77	+ createdAt: 2026-10-21T08:00:00.000Z
	78	+ updatedAt: 2026-10-21T08:00:00.000Z
	79	+ - id: 3
	80	+ title: "Docs EN+FR: reference/tools.md plan › Storage row"
	81	+ state: closed
	82	+ priority: none
	83	+ author:
	84	+ name: k33g
	85	+ email: ph.charriere@gmail.com
	86	+ createdAt: 2026-10-21T08:00:00.000Z
	87	+ updatedAt: 2026-10-21T08:00:00.000Z
	88	+comments:
	89	+ - id: 1
	90	+ author:
	91	+ name: k33g
	92	+ email: ph.charriere@gmail.com
	93	+ createdAt: 2026-10-21T10:00:00.000Z
	94	+ body: \|
	95	+ Fixed on fix/plan-mode (user's request: "fixe les 3 tickets").
	96	+ - internal/ui: optional interface Conversation { ConversationID() }, and
	97	+ ui.ConversationID() — the active sink's, "" for the front ends that hold
	98	+ one conversation (line mode, window).
	99	+ - internal/acp/sink.go: ConversationID() = the ACP session id.
	100	+ - internal/tools/plan.go: getSessionID() = ui.ConversationID() (was the
	101	+ constant "current-session"); ClearPlans(conversation) replaces
	102	+ ClearAllPlans().
	103	+ - Tests: tools/plan_conversation_test.go TestPlansArePerConversation (two
	104	+ threads: each first unnamed plan is user-request-1, A never lists or
	105	+ reads B's plan, clearing A keeps B's).
	106	+ - Docs EN+FR: reference/tools.md plan › Storage.

++# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
++id: 108
++title: "plan: one store for the whole process — every ACP session sees the same plans"
++state: closed
++author:
++  name: k33g
++  email: ph.charriere@gmail.com
++createdAt: 2026-10-21T08:00:00.000Z
++updatedAt: 2026-10-21T10:00:00.000Z
++labels:
++  - kind::bug
++milestone: v0.9.0
++epic: Builtin tools
++related:
++  - 109
++  - 110
++body: |
++  ## The problem
++
++  The `plan` tool keeps its plans in memory, in ONE store for the whole mm
++  process: `internal/tools/plan.go`, `getSessionID()` returns the constant
++  "current-session" ("v1: use a single session for simplicity"). Every plan
++  lives under that one key.
++
++  In the line mode and the window this goes unnoticed: one process, one
++  conversation at a time. Over ACP it does not hold: one `mm -acp` process
++  serves every thread the editor opens (Zed: each thread is a session/new),
++  and they all share that one store. Concretely, with two threads open in
++  Zed:
++
++  - thread A asks "do these 3 things…": the model creates `user-request-1`;
++  - thread B, about something else, asks for a plan too: its unnamed plan
++    becomes `user-request-2` (the name counts A's plan), and
++    `plan list` in B answers with BOTH plans;
++  - the model in B can read, `update`, `next` or `delete` A's plan — a small
++    model told "Found 2 plan(s)" will happily work on the wrong one;
++  - a `create` with a name A already used fails with
++    `plan 'x' already exists`, for a thread that never created it.
++
++  Nothing tells the user or the model that the plans come from another
++  conversation. It is also why plans cannot be written to the session file
++  (the code's own "v2" note): they are not attached to a session at all.
++
++  ## What fixing it solves
++
++  - Each conversation has its own plans: one Zed thread can no longer see,
++    change or delete another thread's plan, and plan names restart at
++    `user-request-1` in each.
++  - A prerequisite for the two follow-ups: /new clearing the right plans in
++    every front end (ticket 109), and later keeping plans in the session file
++    so /resume and session/load bring them back.
++
++  ## Direction (to confirm)
++
++  Key the store by the conversation the call belongs to. The tool does not
++  know it today: the session id is not passed to tools (the same gap
++  search_memory worked around in ticket 48). Options: a session id in the
++  tool's context, set by each front end next to ui.SetActive; or one plan
++  store per front-end session, handed to the tool through the active sink.
++tasks:
++  - id: 1
++    title: "Choose how a tool call learns its conversation (context value vs the active sink) — the same gap as ticket 48's search_memory"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++  - id: 2
++    title: "Key the plan store by that conversation; plans of one ACP session invisible to another — with a test of two sessions over the wire"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++  - id: 3
++    title: "Docs EN+FR: reference/tools.md plan › Storage row"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++comments:
++  - id: 1
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T10:00:00.000Z
++    body: |
++      Fixed on fix/plan-mode (user's request: "fixe les 3 tickets").
++      - internal/ui: optional interface Conversation { ConversationID() }, and
++        ui.ConversationID() — the active sink's, "" for the front ends that hold
++        one conversation (line mode, window).
++      - internal/acp/sink.go: ConversationID() = the ACP session id.
++      - internal/tools/plan.go: getSessionID() = ui.ConversationID() (was the
++        constant "current-session"); ClearPlans(conversation) replaces
++        ClearAllPlans().
++      - Tests: tools/plan_conversation_test.go TestPlansArePerConversation (two
++        threads: each first unnamed plan is user-request-1, A never lists or
++        reads B's plan, clearing A keeps B's).
++      - Docs EN+FR: reference/tools.md plan › Storage.

added .tickets/issues/0109-new-does-not-clear-plans-in-window-and-acp.yaml +99 -0 Viewed

new file mode 100644
@@ -0,0 +1,99 @@
	1	+# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
	2	+id: 109
	3	+title: "/new clears the plans in the line mode only — not in the window, not over ACP"
	4	+state: closed
	5	+author:
	6	+ name: k33g
	7	+ email: ph.charriere@gmail.com
	8	+createdAt: 2026-10-21T08:00:00.000Z
	9	+updatedAt: 2026-10-21T10:00:00.000Z
	10	+labels:
	11	+ - kind::bug
	12	+milestone: v0.9.0
	13	+epic: Builtin tools
	14	+related:
	15	+ - 108
	16	+ - 110
	17	+body: \|
	18	+ ## The problem
	19	+
	20	+ `/new` means "a new session: forget the conversation". In the line mode it
	21	+ also forgets the plans: `internal/agent/agent.go` startNewSession calls
	22	+ `tools.ClearAllPlans()`. The two other front ends do not:
	23	+
	24	+ - the window: `internal/tui/commands.go` runner.newSession resets the
	25	+ history, the token count, the loop detector and the session file — not
	26	+ the plans;
	27	+ - ACP: `internal/acp/acp.go` front.startNewSession archives the
	28	+ conversation and resets the history — not the plans.
	29	+
	30	+ So after `/new` in the window or in an editor, the conversation is empty
	31	+ but the plan of the previous task is still there: the next `plan list`
	32	+ returns it, an unnamed `create` is called `user-request-2`, and a model that
	33	+ reads the old plan may go back to tasks of a conversation the user just
	34	+ closed. The same `/new` behaves differently depending on where it is typed
	35	+ — the line mode is right, the two others are not.
	36	+
	37	+ (Over ACP the call would today clear EVERY session's plans, since the store
	38	+ is shared — ticket 108. The two fixes go together.)
	39	+
	40	+ ## What fixing it solves
	41	+
	42	+ - `/new` forgets the plans in the three front ends, as it forgets the
	43	+ history: a new task starts with no leftover plan.
	44	+ - One rule for all front ends, as the shared command table promises.
	45	+
	46	+ ## Direction (to confirm)
	47	+
	48	+ Call the plan reset from the shared new-session path (or from each of the
	49	+ two missing ones), and — once ticket 108 is done — reset only the plans of
	50	+ the session being renewed. Decide too what /resume should do with the
	51	+ plans in memory (today: nothing, in every front end).
	52	+tasks:
	53	+ - id: 1
	54	+ title: "Clear the plans on /new in the window and over ACP (after or with ticket 108, so an ACP /new clears only its own session) — with tests"
	55	+ state: closed
	56	+ priority: none
	57	+ author:
	58	+ name: k33g
	59	+ email: ph.charriere@gmail.com
	60	+ createdAt: 2026-10-21T08:00:00.000Z
	61	+ updatedAt: 2026-10-21T08:00:00.000Z
	62	+ - id: 2
	63	+ title: "Decide and document what /resume and session/load do with the plans in memory"
	64	+ state: closed
	65	+ priority: none
	66	+ author:
	67	+ name: k33g
	68	+ email: ph.charriere@gmail.com
	69	+ createdAt: 2026-10-21T08:00:00.000Z
	70	+ updatedAt: 2026-10-21T08:00:00.000Z
	71	+ - id: 3
	72	+ title: "Docs EN+FR: reference/tools.md plan › Storage row; reference/slash-commands.md /new"
	73	+ state: closed
	74	+ priority: none
	75	+ author:
	76	+ name: k33g
	77	+ email: ph.charriere@gmail.com
	78	+ createdAt: 2026-10-21T08:00:00.000Z
	79	+ updatedAt: 2026-10-21T08:00:00.000Z
	80	+comments:
	81	+ - id: 1
	82	+ author:
	83	+ name: k33g
	84	+ email: ph.charriere@gmail.com
	85	+ createdAt: 2026-10-21T10:00:00.000Z
	86	+ body: \|
	87	+ Fixed on fix/plan-mode. Decision (task 2): plans belong to a conversation
	88	+ and are not saved, so a conversation that changes starts with no plan:
	89	+ - line mode: /new (already) and /resume → tools.ClearPlans("");
	90	+ - window: runner.newSession and runner.resume → tools.ClearPlans("");
	91	+ - ACP: startNewSession → tools.ClearPlans(sid) (this thread only);
	92	+ session/load and session/resume (the shared restore) →
	93	+ tools.ClearPlans(id).
	94	+ - Tests: tui TestRunnerNewAndResumeClearThePlans (fails without the fix:
	95	+ checked by removing the two calls), acp
	96	+ TestNewOverACPClearsOnlyItsThreadsPlans (over the wire, two sessions),
	97	+ acp TestLoadSessionStartsWithNoPlan.
	98	+ - Docs EN+FR: reference/tools.md plan › Storage; slash-commands /new
	99	+ (terminal + ACP) and /resume.

++# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
++id: 109
++title: "/new clears the plans in the line mode only — not in the window, not over ACP"
++state: closed
++author:
++  name: k33g
++  email: ph.charriere@gmail.com
++createdAt: 2026-10-21T08:00:00.000Z
++updatedAt: 2026-10-21T10:00:00.000Z
++labels:
++  - kind::bug
++milestone: v0.9.0
++epic: Builtin tools
++related:
++  - 108
++  - 110
++body: |
++  ## The problem
++
++  `/new` means "a new session: forget the conversation". In the line mode it
++  also forgets the plans: `internal/agent/agent.go` startNewSession calls
++  `tools.ClearAllPlans()`. The two other front ends do not:
++
++  - the window: `internal/tui/commands.go` runner.newSession resets the
++    history, the token count, the loop detector and the session file — not
++    the plans;
++  - ACP: `internal/acp/acp.go` front.startNewSession archives the
++    conversation and resets the history — not the plans.
++
++  So after `/new` in the window or in an editor, the conversation is empty
++  but the plan of the previous task is still there: the next `plan list`
++  returns it, an unnamed `create` is called `user-request-2`, and a model that
++  reads the old plan may go back to tasks of a conversation the user just
++  closed. The same `/new` behaves differently depending on where it is typed
++  — the line mode is right, the two others are not.
++
++  (Over ACP the call would today clear EVERY session's plans, since the store
++  is shared — ticket 108. The two fixes go together.)
++
++  ## What fixing it solves
++
++  - `/new` forgets the plans in the three front ends, as it forgets the
++    history: a new task starts with no leftover plan.
++  - One rule for all front ends, as the shared command table promises.
++
++  ## Direction (to confirm)
++
++  Call the plan reset from the shared new-session path (or from each of the
++  two missing ones), and — once ticket 108 is done — reset only the plans of
++  the session being renewed. Decide too what /resume should do with the
++  plans in memory (today: nothing, in every front end).
++tasks:
++  - id: 1
++    title: "Clear the plans on /new in the window and over ACP (after or with ticket 108, so an ACP /new clears only its own session) — with tests"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++  - id: 2
++    title: "Decide and document what /resume and session/load do with the plans in memory"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++  - id: 3
++    title: "Docs EN+FR: reference/tools.md plan › Storage row; reference/slash-commands.md /new"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++comments:
++  - id: 1
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T10:00:00.000Z
++    body: |
++      Fixed on fix/plan-mode. Decision (task 2): plans belong to a conversation
++      and are not saved, so a conversation that changes starts with no plan:
++      - line mode: /new (already) and /resume → tools.ClearPlans("");
++      - window: runner.newSession and runner.resume → tools.ClearPlans("");
++      - ACP: startNewSession → tools.ClearPlans(sid) (this thread only);
++        session/load and session/resume (the shared restore) →
++        tools.ClearPlans(id).
++      - Tests: tui TestRunnerNewAndResumeClearThePlans (fails without the fix:
++        checked by removing the two calls), acp
++        TestNewOverACPClearsOnlyItsThreadsPlans (over the wire, two sessions),
++        acp TestLoadSessionStartsWithNoPlan.
++      - Docs EN+FR: reference/tools.md plan › Storage; slash-commands /new
++        (terminal + ACP) and /resume.

added .tickets/issues/0110-plans-how-to-promises-progress-lines-mm-never-prints.yaml +134 -0 Viewed

new file mode 100644
@@ -0,0 +1,134 @@
	1	+# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
	2	+id: 110
	3	+title: "Plans how-to shows progress lines (📋 [2/3] …) that mm never prints"
	4	+state: closed
	5	+author:
	6	+ name: k33g
	7	+ email: ph.charriere@gmail.com
	8	+createdAt: 2026-10-21T08:00:00.000Z
	9	+updatedAt: 2026-10-21T10:00:00.000Z
	10	+labels:
	11	+ - kind::bug
	12	+milestone: v0.9.0
	13	+epic: Documentation
	14	+related:
	15	+ - 108
	16	+ - 109
	17	+body: \|
	18	+ ## The problem
	19	+
	20	+ docs/en/how-to/manage-plans-and-tasks.md (and its French twin) promises
	21	+ that mm shows where the agent is in a plan "after each response", with
	22	+ lines like:
	23	+
	24	+ 📋 [2/3] 🔄 Task 2: Add a license (in_progress)
	25	+ 📋 [3/3] ✅ All tasks complete!
	26	+
	27	+ and builds most of the page on them ("The 📋 line shows you exactly where
	28	+ the agent is in your plan", worked examples, a "status line format"
	29	+ section). No code prints those lines. What mm actually does
	30	+ (internal/tools/plan.go):
	31	+
	32	+ - line mode: it prints ONE line before each call, `📋 plan: <action>
	33	+ <name>` (e.g. `📋 plan: update user-request-1`), and nothing after: the
	34	+ result is not echoed;
	35	+ - the result — JSON text holding a `status` field such as
	36	+ `[2/3] 🔄 Task 2: Add a license (in_progress)` — goes to the MODEL. The
	37	+ tool's description asks the model to "include [it] in your response to
	38	+ the user"; a model that does shows a similar line, without the 📋, in its
	39	+ own words; one that does not shows nothing;
	40	+ - window and editor: the call appears as a tool call, its raw JSON result
	41	+ in the tool's frame.
	42	+
	43	+ The page also announces a `/plans` command ("not yet implemented in v1 —
	44	+ coming in v2") with a sample output, in a section titled "Plan Management
	45	+ Commands".
	46	+
	47	+ So a user who reads the guide expects a progress line mm never draws, and
	48	+ cannot tell whether a missing line means a bug, an old version, or a model
	49	+ that skipped it.
	50	+
	51	+ ## What fixing it solves
	52	+
	53	+ - The guide describes what the user will see, front end by front end, and
	54	+ says plainly that progress shown in the answer depends on the model.
	55	+ - If progress lines ARE wanted, that is a feature to decide (mm printing
	56	+ the status after each plan call, or after each answer), not something
	57	+ the docs pretend exists.
	58	+
	59	+ ## Direction (to decide)
	60	+
	61	+ Either rewrite the how-to (EN+FR) to match the code — the reference
	62	+ section written on 2026-10-21 already does —, or implement the promised
	63	+ line (e.g. print the `status` line after each plan action in the line mode
	64	+ and as a notice in the window) and then document it. The `/plans` section
	65	+ should go, or become a ticket of its own.
	66	+tasks:
	67	+ - id: 1
	68	+ title: "Decide: rewrite the how-to to match mm, or implement a progress line mm prints itself"
	69	+ state: closed
	70	+ priority: none
	71	+ author:
	72	+ name: k33g
	73	+ email: ph.charriere@gmail.com
	74	+ createdAt: 2026-10-21T08:00:00.000Z
	75	+ updatedAt: 2026-10-21T08:00:00.000Z
	76	+ - id: 2
	77	+ title: "Docs EN+FR: manage-plans-and-tasks.md matches what each front end shows; the /plans section removed or turned into a ticket"
	78	+ state: closed
	79	+ priority: none
	80	+ author:
	81	+ name: k33g
	82	+ email: ph.charriere@gmail.com
	83	+ createdAt: 2026-10-21T08:00:00.000Z
	84	+ updatedAt: 2026-10-21T08:00:00.000Z
	85	+comments:
	86	+ - id: 1
	87	+ author:
	88	+ name: k33g
	89	+ email: ph.charriere@gmail.com
	90	+ createdAt: 2026-10-21T10:00:00.000Z
	91	+ body: \|
	92	+ Fixed on fix/plan-mode. Decision (task 1): implement the line, then
	93	+ document it.
	94	+ - internal/tools/plan.go progressLine: after every plan action whose result
	95	+ has a status (create — which now returns one —, update, next, status) mm
	96	+ shows `📋 <status line>`: printed in the line mode, a notice in the window
	97	+ (ui.Noticer); an editor shows the status in the tool call's result.
	98	+ - Making it visible exposed a pre-existing defect (live run, Qwen3.5 4B):
	99	+ the status line followed plan.Current, which only `next` moved, so after
	100	+ `update 1 done` it still showed task 1, and a later `next` put the
	101	+ finished task 1 back in progress. Fixed here, since the progress shown
	102	+ must be right: the line shows the first task in progress, else the first
	103	+ open one (`Plan finished: D done, F failed` when over with failures);
	104	+ `update … in_progress` makes the task current; `next` skips finished
	105	+ tasks. TestRunPlanAction encoded the old behaviour (next after update 1
	106	+ in_progress restarted task 1) — expectation corrected; quote escaping
	107	+ kept in its own test.
	108	+ - Cosmetic: the call title of an unnamed plan no longer ends with a space
	109	+ (planTitle).
	110	+ - Live (mm -p, Qwen3.5 4B): 📋 [1/2] … open → 📋 [2/2] … open → 📋 [2/2]
	111	+ All tasks complete!, files written.
	112	+ - Docs EN+FR: how-to manage-plans-and-tasks rewritten (450 → 64 lines: what
	113	+ each front end shows, the measured sample, steering in words, variants,
	114	+ troubleshooting; the /plans section removed — no such command), now in
	115	+ both indexes; reference/tools.md plan (create/update/next/status line/
	116	+ display rows); root README Plans line.
	117	+ - id: 2
	118	+ author:
	119	+ name: k33g
	120	+ email: ph.charriere@gmail.com
	121	+ createdAt: 2026-10-21T11:00:00.000Z
	122	+ body: \|
	123	+ Follow-up, found by demos/plans (user's request: a plans demo): Qwen3.5
	124	+ 4B never passed `name` after create; every update failed ("no plan
	125	+ named ''") and the turn ended in runaway tool calls. An action without
	126	+ a name now works on the conversation's most recently created plan
	127	+ (PlanStore.LatestName, by creation order — CreatedAt has one-second
	128	+ precision); the field's description said "Defaults to 'user-request'",
	129	+ a name no plan has. Test TestPlanActionWithoutANameUsesTheLatestPlan.
	130	+ Docs EN+FR: reference plan › name. demos/plans measured: 4-step Go
	131	+ program 3/3 (go test passes), failing push 3/3 "Plan finished: 2 done,
	132	+ 1 failed" once the prompt forbids stand-ins (before: marked done 3/3; in
	133	+ the run inspected, a bare repo created in /tmp and force-pushed).
	134	+

++# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
++id: 110
++title: "Plans how-to shows progress lines (📋 [2/3] …) that mm never prints"
++state: closed
++author:
++  name: k33g
++  email: ph.charriere@gmail.com
++createdAt: 2026-10-21T08:00:00.000Z
++updatedAt: 2026-10-21T10:00:00.000Z
++labels:
++  - kind::bug
++milestone: v0.9.0
++epic: Documentation
++related:
++  - 108
++  - 109
++body: |
++  ## The problem
++
++  docs/en/how-to/manage-plans-and-tasks.md (and its French twin) promises
++  that mm shows where the agent is in a plan "after each response", with
++  lines like:
++
++      📋 [2/3] 🔄 Task 2: Add a license (in_progress)
++      📋 [3/3] ✅ All tasks complete!
++
++  and builds most of the page on them ("The 📋 line shows you exactly where
++  the agent is in your plan", worked examples, a "status line format"
++  section). No code prints those lines. What mm actually does
++  (internal/tools/plan.go):
++
++  - line mode: it prints ONE line before each call, `📋 plan: <action>
++    <name>` (e.g. `📋 plan: update user-request-1`), and nothing after: the
++    result is not echoed;
++  - the result — JSON text holding a `status` field such as
++    `[2/3] 🔄 Task 2: Add a license (in_progress)` — goes to the MODEL. The
++    tool's description asks the model to "include [it] in your response to
++    the user"; a model that does shows a similar line, without the 📋, in its
++    own words; one that does not shows nothing;
++  - window and editor: the call appears as a tool call, its raw JSON result
++    in the tool's frame.
++
++  The page also announces a `/plans` command ("not yet implemented in v1 —
++  coming in v2") with a sample output, in a section titled "Plan Management
++  Commands".
++
++  So a user who reads the guide expects a progress line mm never draws, and
++  cannot tell whether a missing line means a bug, an old version, or a model
++  that skipped it.
++
++  ## What fixing it solves
++
++  - The guide describes what the user will see, front end by front end, and
++    says plainly that progress shown in the answer depends on the model.
++  - If progress lines ARE wanted, that is a feature to decide (mm printing
++    the status after each plan call, or after each answer), not something
++    the docs pretend exists.
++
++  ## Direction (to decide)
++
++  Either rewrite the how-to (EN+FR) to match the code — the reference
++  section written on 2026-10-21 already does —, or implement the promised
++  line (e.g. print the `status` line after each plan action in the line mode
++  and as a notice in the window) and then document it. The `/plans` section
++  should go, or become a ticket of its own.
++tasks:
++  - id: 1
++    title: "Decide: rewrite the how-to to match mm, or implement a progress line mm prints itself"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++  - id: 2
++    title: "Docs EN+FR: manage-plans-and-tasks.md matches what each front end shows; the /plans section removed or turned into a ticket"
++    state: closed
++    priority: none
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T08:00:00.000Z
++    updatedAt: 2026-10-21T08:00:00.000Z
++comments:
++  - id: 1
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T10:00:00.000Z
++    body: |
++      Fixed on fix/plan-mode. Decision (task 1): implement the line, then
++      document it.
++      - internal/tools/plan.go progressLine: after every plan action whose result
++        has a status (create — which now returns one —, update, next, status) mm
++        shows `📋 <status line>`: printed in the line mode, a notice in the window
++        (ui.Noticer); an editor shows the status in the tool call's result.
++      - Making it visible exposed a pre-existing defect (live run, Qwen3.5 4B):
++        the status line followed plan.Current, which only `next` moved, so after
++        `update 1 done` it still showed task 1, and a later `next` put the
++        finished task 1 back in progress. Fixed here, since the progress shown
++        must be right: the line shows the first task in progress, else the first
++        open one (`Plan finished: D done, F failed` when over with failures);
++        `update … in_progress` makes the task current; `next` skips finished
++        tasks. TestRunPlanAction encoded the old behaviour (next after update 1
++        in_progress restarted task 1) — expectation corrected; quote escaping
++        kept in its own test.
++      - Cosmetic: the call title of an unnamed plan no longer ends with a space
++        (planTitle).
++      - Live (mm -p, Qwen3.5 4B): 📋 [1/2] … open → 📋 [2/2] … open → 📋 [2/2]
++        All tasks complete!, files written.
++      - Docs EN+FR: how-to manage-plans-and-tasks rewritten (450 → 64 lines: what
++        each front end shows, the measured sample, steering in words, variants,
++        troubleshooting; the /plans section removed — no such command), now in
++        both indexes; reference/tools.md plan (create/update/next/status line/
++        display rows); root README Plans line.
++  - id: 2
++    author:
++      name: k33g
++      email: ph.charriere@gmail.com
++    createdAt: 2026-10-21T11:00:00.000Z
++    body: |
++      Follow-up, found by demos/plans (user's request: a plans demo): Qwen3.5
++      4B never passed `name` after create; every update failed ("no plan
++      named ''") and the turn ended in runaway tool calls. An action without
++      a name now works on the conversation's most recently created plan
++      (PlanStore.LatestName, by creation order — CreatedAt has one-second
++      precision); the field's description said "Defaults to 'user-request'",
++      a name no plan has. Test TestPlanActionWithoutANameUsesTheLatestPlan.
++      Docs EN+FR: reference plan › name. demos/plans measured: 4-step Go
++      program 3/3 (go test passes), failing push 3/3 "Plan finished: 2 done,
++      1 failed" once the prompt forbids stand-ins (before: marked done 3/3; in
++      the run inspected, a bare repo created in /tmp and force-pushed).
++

added .tickets/issues/0111-using-the-settings-guide-create-a-claude-skill.yaml +15 -0 Viewed

new file mode 100644
@@ -0,0 +1,15 @@
	1	+# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
	2	+id: 111
	3	+title: Using the settings guide create a Claude skill
	4	+state: open
	5	+author:
	6	+ name: k33g
	7	+ email: ph.charriere@gmail.com
	8	+createdAt: 2026-10-21T04:27:19.433Z
	9	+updatedAt: 2026-10-21T04:41:59.368Z
	10	+labels: []
	11	+milestone: v1.0.0
	12	+body: \|
	13	+ the Claude skill will produce a descriptive doc about the model and a config file (`agent.yaml`)
	14	+tasks: []
	15	+comments: []

++# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
++id: 111
++title: Using the settings guide create a Claude skill
++state: open
++author:
++  name: k33g
++  email: ph.charriere@gmail.com
++createdAt: 2026-10-21T04:27:19.433Z
++updatedAt: 2026-10-21T04:41:59.368Z
++labels: []
++milestone: v1.0.0
++body: |
++  the Claude skill will produce a descriptive doc about the model and a config file (`agent.yaml`)
++tasks: []
++comments: []

added .tickets/issues/0112-add-a-help-window-in-tui-mode-for-the-ui-shortcu.yaml +15 -0 Viewed

new file mode 100644
@@ -0,0 +1,15 @@
	1	+# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
	2	+id: 112
	3	+title: Add a help window in TUI mode for the UI shortcuts
	4	+state: open
	5	+author:
	6	+ name: k33g
	7	+ email: ph.charriere@gmail.com
	8	+createdAt: 2026-10-21T04:43:03.511Z
	9	+updatedAt: 2026-10-21T04:43:13.927Z
	10	+labels:
	11	+ - kind::feature
	12	+milestone: v0.9.0
	13	+body: ""
	14	+tasks: []
	15	+comments: []

++# Managed by IssueSpec. Hand edits are welcome; keep the schema valid.
++id: 112
++title: Add a help window in TUI mode for the UI shortcuts
++state: open
++author:
++  name: k33g
++  email: ph.charriere@gmail.com
++createdAt: 2026-10-21T04:43:03.511Z
++updatedAt: 2026-10-21T04:43:13.927Z
++labels:
++  - kind::feature
++milestone: v0.9.0
++body: ""
++tasks: []
++comments: []

modified README.md +6 -4 Viewed

@@ -65,7 +65,7 @@ Type a question at the `>` prompt. Each command the model runs is shown with its
65	65	mm -tui
66	66	```
67	67
68		-The conversation above, a box to type in below. Enter sends, Ctrl-J adds a line, Esc cancels a turn, F1 lists the keys. With `yolo: false` in the YAML, the agent asks before running a command or writing a file — in the window, the line mode and an editor alike (`yolo: true`, the default, runs them unasked).
	68	+The conversation above, a box to type in below. Enter sends, Ctrl-J adds a line, Esc cancels a turn (or clears the box when nothing runs), F1 lists the keys. With sessions on, ↑ in the box brings back your earlier questions, from this session and the project's previous ones. With `yolo: false` in the YAML, the agent asks before running a command or writing a file — in the window, the line mode and an editor alike (`yolo: true`, the default, runs them unasked).
69	69
70	70	### 3. Inside an editor (ACP)
71	71
@@ -96,7 +96,7 @@ Speaks the Agent Client Protocol on stdin/stdout. You never run this one by hand
96	96	`AGENT_CONFIG` must be an absolute path: Zed starts `mm` in the project directory, and a relative path would not be found. `"type": "custom"` is required.
97	97
98	98	3. Open the Agent panel, pick mini-me, and talk to it. Every `bash`, `read_file`, `write_file` and `edit_file` call shows up as a tool call, and file changes appear as real diffs. With `yolo: false` in the YAML, Zed asks you to allow each call before it runs; the shipped files say `yolo: true`, and the calls run unasked.
99		-4. Handy things: Escape cancels a turn; `/new` and `/compact` are in the slash-command menu; `@file` attaches a file for the model to read; with `sessions.enabled: true` in the YAML, Reload Agent brings the whole conversation back, and the Agent panel's thread history lists the project's past conversations so you can reopen one (it is not in the `/` menu).
	99	+4. Handy things: Escape cancels a turn; `/new` and `/compact` are in the slash-command menu (with `/memory` and `/docs` when you turn them on); `@file` attaches a file for the model to read; with `sessions.enabled: true` in the YAML, Reload Agent brings the whole conversation back, and the Agent panel's thread history lists the project's past conversations so you can reopen one (it is not in the `/` menu).
100	100	5. When something looks wrong, open `dev: open acp logs`: the banner, warnings and the agent's `[acp] …` trail go there.
101	101
102	102	![mm-acp.png](/bots-garden/mini-me/raw/main/assets/mm-acp.png)
@@ -114,11 +114,11 @@ Anything that speaks ACP can host `mm -acp` the same way: JetBrains IDEs (In
114	114	- Project instructions: an `AGENTS.md` in the project directory is added to the system prompt of every conversation. See [Give the agent your project's instructions](docs/en/how-to/give-project-instructions.md).
115	115	- Answers as JSON: declare the shape of the answer as Go types under `structuredOutput`, and `mm -p "…" \| jq` works. See [Get answers as JSON](docs/en/how-to/get-json-answers.md).
116	116	- MCP servers: their tools, wrapped as mm tools. See the [MCP tutorial](docs/en/tutorials/mcp-getting-started.md).
117		-- Sessions: with `sessions.enabled: true`, `/sessions` lists the conversations of a project and `/resume` continues one; in an editor, its thread history lists them (`session/list`) and reopens one (`session/load`, `session/resume`). See [Keep sessions and resume them](docs/en/how-to/keep-sessions.md).
	117	+- Sessions: with `sessions.enabled: true`, `/sessions` lists the conversations of a project and `/resume` continues one; in the window, ↑ recalls your earlier questions to send them again; in an editor, its thread history lists them (`session/list`) and reopens one (`session/load`, `session/resume`). See [Keep sessions and resume them](docs/en/how-to/keep-sessions.md).
118	118	- Trace: `AGENT_TRACE=true mm …` writes every request to the model and every response to `.mm/traces/`. See [See the JSON sent to the model](docs/en/how-to/inspect-model-traffic.md).
119	119	- Hooks: your own commands — in any language — run before a question reaches the model, before and after each tool call, after the turn, and around each hand-off to another agent; they can deny a call, stop or rewrite a question, add context, mask what leaves for another agent, replace what comes back, or log. `displayHooks: true` shows each hook that ran and what it decided, after the answer (an editor shows each one as a tool call). See [Guard the agent with hooks](docs/en/how-to/guard-the-agent-with-hooks.md), [Control the hand-offs to other agents with hooks](docs/en/how-to/control-agent-calls-with-hooks.md), `demos/hooks/` and `demos/hooks-ask/`.
120	120	- Guard models: have every question checked by guard models (IBM's Granite Guardian), one agent per risk, before your assistant answers — a `preGeneration` hook asks them without the conversation history, and the assistant refuses or answers according to their verdicts. See [Screen questions with guard models](docs/en/how-to/screen-questions-with-guard-models.md), `demos/guard/` and `demos/guards/`.
121		-- Plans: the built-in `plan` tool lets the agent turn a list of tasks into a tracked plan and show its progress after each answer. See [Manage plans and tasks](docs/en/how-to/manage-plans-and-tasks.md).
	121	+- Plans: the built-in `plan` tool lets the agent turn a list of tasks into a tracked plan; mm shows its progress (`📋 [2/3] …`) after each plan action — per conversation, cleared by `/new` and `/resume`. See [Manage plans and tasks](docs/en/how-to/manage-plans-and-tasks.md).
122	122	- Save an answer: `/save` writes the last answer to a Markdown file; with `saveAnswerTool: true`, the model does it itself when you ask in your own words (`save_answer`). Never overwrites a file. See [Save an answer to a file](docs/en/how-to/save-an-answer.md).
123	123	- Unload a model: `/unload [model]` frees a model from the server's memory — in the terminal, the window and an editor —, and the `unload_model` tool lets the agent do it itself, e.g. on another agent's model once it has answered; llmman, Ollama and llama-server in router mode. `unloadModels: false` turns both off. See [Free a model's memory](docs/en/how-to/unload-a-model.md) and `demos/unload-models/`.
124	124	- Search a documentation folder: `/docs <words>` searches the text files of the folders named in `docSearch.dirs` (`.md`, `.adoc`, `.txt`… — only the extensions listed) and returns the best passages with their `path:line` link, ranked with BM25 — no index, no model (`docSearch.searchDocs`); the `search_docs` tool lets the model do it before it answers (`docSearch.autoSearchDocs`); both off by default. See [Search a documentation folder](docs/en/how-to/search-the-docs.md) and `demos/docs-search/`.
@@ -156,6 +156,7 @@ The `demos/` directory holds ready-made set-ups (most with a README explaining h
156	156	\| `demos/guard/` \| one guard model (Granite Guardian 3.2 3B-A800M on llmman) answering `Yes`/`No` to "is this question harmful?"; why its `system:` line is a risk name, not a prompt; custom risks \|
157	157	\| `demos/guards/` \| an assistant (Gemma 4 E4B) protected by seven guard agents, one per risk: a Node.js `preGeneration` hook asks them in parallel without the history, the assistant refuses or answers according to their verdicts; `displayHooks: true`; measured results \|
158	158	\| `demos/unload-models/` \| a coordinator (Qwen3.5 9B) hands programming questions to an expert (Qwen3.5 4B) with `ask_agent`, a `preAsk` hook unloads the coordinator's own model before the expert starts and a `postAsk` hook unloads the expert's once it has answered — the two models are never loaded together; `/unload` and `unload_model` on request; llmman, measured results \|
	159	+\| `demos/plans/` \| the `plan` tool at work: Planner (Qwen3.5 4B on llmman) builds a tiny Go program in four steps — module, code, test, README — while mm shows `📋 [C/N] …` after each plan action; a failing variant ends on `Plan finished: 2 done, 1 failed`; measured results \|
159	160	\| `demos/docs-search/` \| an assistant answering from a documentation folder: nine Ratatui tutorials (Rust terminal UIs, each program checked against ratatui 0.30 by `check-tutorials.sh`), `/docs <words>` for you and the `search_docs` tool for the model; Qwen3.5 4B on llmman, measured results \|
160	161	\| `demos/memory-search/` \| an assistant that remembers a project's decisions: four earlier sessions (port, storage, logs in French, deployment as a `/compact` summary), `/memory <words>` for you and the `search_memory` tool for the model; Qwen3.5 4B on llmman, measured results — including what a word search cannot find \|
161	162	\| `demos/declared-tools/` \| an ops desk with no `bash` and no skills, acting only through four tools declared under `tools:` — release, version bump, service logs, slow requests — covering every parameter type (string, enum, boolean, integer, number), a safe default (`dryRun: true`) and an optional argument; Mellum2 12B on DMR, measured results \|
@@ -163,6 +164,7 @@ The `demos/` directory holds ready-made set-ups (most with a README explaining h
163	164	## Learn more
164	165
165	166	- [Documentation](docs/README.md) — tutorials, how-to guides, reference and explanations, in English and French.
	167	+- [Release notes](release-notes/README.md) — what changed in each version, in English and French.
166	168	- [`guides/`](guides/) — practical guides; for now, the agent settings for small models, in [English](guides/agent-settings-for-small-models.md) and [French](guides/reglages-agents-petits-modeles.md): timeouts, `maxTurns`, `maxOutput`, `agentsMd.maxChars`, context compression, sampling — what each one does, its default, and how to set it for a small local model.
167	169	- The `skills/` directory: Markdown procedures the agent can load with its `read_skill` tool.
168	170	- `agent.lsp.yaml`: the same agent with a language server behind an `lsp` tool (definition, references, hover, symbols, diagnostics, rename) — see [Navigate and refactor code with a language server](docs/en/how-to/navigate-code-with-lsp.md).

  mm -tui
  ```
--The conversation above, a box to type in below. **Enter** sends, **Ctrl-J** adds a line, **Esc** cancels a turn, **F1** lists the keys. With `yolo: false` in the YAML, the agent asks before running a command or writing a file — in the window, the line mode and an editor alike (`yolo: true`, the default, runs them unasked).
++The conversation above, a box to type in below. **Enter** sends, **Ctrl-J** adds a line, **Esc** cancels a turn (or clears the box when nothing runs), **F1** lists the keys. With sessions on, **↑** in the box brings back your earlier questions, from this session and the project's previous ones. With `yolo: false` in the YAML, the agent asks before running a command or writing a file — in the window, the line mode and an editor alike (`yolo: true`, the default, runs them unasked).
  ### 3. Inside an editor (ACP)
     `AGENT_CONFIG` must be an **absolute** path: Zed starts `mm` in the project directory, and a relative path would not be found. `"type": "custom"` is required.
 . Open the Agent panel, pick **mini-me**, and talk to it. Every `bash`, `read_file`, `write_file` and `edit_file` call shows up as a tool call, and file changes appear as real diffs. With `yolo: false` in the YAML, Zed asks you to allow each call before it runs; the shipped files say `yolo: true`, and the calls run unasked.
--4. Handy things: **Escape** cancels a turn; `/new` and `/compact` are in the slash-command menu; `@file` attaches a file for the model to read; with `sessions.enabled: true` in the YAML, **Reload Agent** brings the whole conversation back, and the Agent panel's thread history lists the project's past conversations so you can reopen one (it is not in the `/` menu).
++4. Handy things: **Escape** cancels a turn; `/new` and `/compact` are in the slash-command menu (with `/memory` and `/docs` when you turn them on); `@file` attaches a file for the model to read; with `sessions.enabled: true` in the YAML, **Reload Agent** brings the whole conversation back, and the Agent panel's thread history lists the project's past conversations so you can reopen one (it is not in the `/` menu).
 . When something looks wrong, open `dev: open acp logs`: the banner, warnings and the agent's `[acp] …` trail go there.
  ![mm-acp.png](/bots-garden/mini-me/raw/main/assets/mm-acp.png)
  - **Project instructions**: an `AGENTS.md` in the project directory is added to the system prompt of every conversation. See [Give the agent your project's instructions](docs/en/how-to/give-project-instructions.md).
  - **Answers as JSON**: declare the shape of the answer as Go types under `structuredOutput`, and `mm -p "…" | jq` works. See [Get answers as JSON](docs/en/how-to/get-json-answers.md).
  - **MCP servers**: their tools, wrapped as mm tools. See the [MCP tutorial](docs/en/tutorials/mcp-getting-started.md).
--- **Sessions**: with `sessions.enabled: true`, `/sessions` lists the conversations of a project and `/resume` continues one; in an editor, its thread history lists them (`session/list`) and reopens one (`session/load`, `session/resume`). See [Keep sessions and resume them](docs/en/how-to/keep-sessions.md).
++- **Sessions**: with `sessions.enabled: true`, `/sessions` lists the conversations of a project and `/resume` continues one; in the window, ↑ recalls your earlier questions to send them again; in an editor, its thread history lists them (`session/list`) and reopens one (`session/load`, `session/resume`). See [Keep sessions and resume them](docs/en/how-to/keep-sessions.md).
  - **Trace**: `AGENT_TRACE=true mm …` writes every request to the model and every response to `.mm/traces/`. See [See the JSON sent to the model](docs/en/how-to/inspect-model-traffic.md).
  - **Hooks**: your own commands — in any language — run before a question reaches the model, before and after each tool call, after the turn, and around each hand-off to another agent; they can deny a call, stop or rewrite a question, add context, mask what leaves for another agent, replace what comes back, or log. `displayHooks: true` shows each hook that ran and what it decided, after the answer (an editor shows each one as a tool call). See [Guard the agent with hooks](docs/en/how-to/guard-the-agent-with-hooks.md), [Control the hand-offs to other agents with hooks](docs/en/how-to/control-agent-calls-with-hooks.md), `demos/hooks/` and `demos/hooks-ask/`.
  - **Guard models**: have every question checked by guard models (IBM's Granite Guardian), one agent per risk, before your assistant answers — a `preGeneration` hook asks them without the conversation history, and the assistant refuses or answers according to their verdicts. See [Screen questions with guard models](docs/en/how-to/screen-questions-with-guard-models.md), `demos/guard/` and `demos/guards/`.
--- **Plans**: the built-in `plan` tool lets the agent turn a list of tasks into a tracked plan and show its progress after each answer. See [Manage plans and tasks](docs/en/how-to/manage-plans-and-tasks.md).
++- **Plans**: the built-in `plan` tool lets the agent turn a list of tasks into a tracked plan; mm shows its progress (`📋 [2/3] …`) after each plan action — per conversation, cleared by `/new` and `/resume`. See [Manage plans and tasks](docs/en/how-to/manage-plans-and-tasks.md).
  - **Save an answer**: `/save` writes the last answer to a Markdown file; with `saveAnswerTool: true`, the model does it itself when you ask in your own words (`save_answer`). Never overwrites a file. See [Save an answer to a file](docs/en/how-to/save-an-answer.md).
  - **Unload a model**: `/unload [model]` frees a model from the server's memory — in the terminal, the window and an editor —, and the `unload_model` tool lets the agent do it itself, e.g. on another agent's model once it has answered; llmman, Ollama and llama-server in router mode. `unloadModels: false` turns both off. See [Free a model's memory](docs/en/how-to/unload-a-model.md) and `demos/unload-models/`.
  - **Search a documentation folder**: `/docs <words>` searches the text files of the folders named in `docSearch.dirs` (`.md`, `.adoc`, `.txt`… — only the extensions listed) and returns the best passages with their `path:line` link, ranked with BM25 — no index, no model (`docSearch.searchDocs`); the `search_docs` tool lets the model do it before it answers (`docSearch.autoSearchDocs`); both off by default. See [Search a documentation folder](docs/en/how-to/search-the-docs.md) and `demos/docs-search/`.
  | `demos/guard/` | one guard model (Granite Guardian 3.2 3B-A800M on llmman) answering `Yes`/`No` to "is this question harmful?"; why its `system:` line is a risk name, not a prompt; custom risks |
  | `demos/guards/` | an assistant (Gemma 4 E4B) protected by seven guard agents, one per risk: a Node.js `preGeneration` hook asks them in parallel without the history, the assistant refuses or answers according to their verdicts; `displayHooks: true`; measured results |
  | `demos/unload-models/` | a coordinator (Qwen3.5 9B) hands programming questions to an expert (Qwen3.5 4B) with `ask_agent`, a `preAsk` hook unloads the coordinator's own model before the expert starts and a `postAsk` hook unloads the expert's once it has answered — the two models are never loaded together; `/unload` and `unload_model` on request; llmman, measured results |
++| `demos/plans/` | the `plan` tool at work: Planner (Qwen3.5 4B on llmman) builds a tiny Go program in four steps — module, code, test, README — while mm shows `📋 [C/N] …` after each plan action; a failing variant ends on `Plan finished: 2 done, 1 failed`; measured results |
  | `demos/docs-search/` | an assistant answering from a documentation folder: nine Ratatui tutorials (Rust terminal UIs, each program checked against ratatui 0.30 by `check-tutorials.sh`), `/docs <words>` for you and the `search_docs` tool for the model; Qwen3.5 4B on llmman, measured results |
  | `demos/memory-search/` | an assistant that remembers a project's decisions: four earlier sessions (port, storage, logs in French, deployment as a `/compact` summary), `/memory <words>` for you and the `search_memory` tool for the model; Qwen3.5 4B on llmman, measured results — including what a word search cannot find |
  | `demos/declared-tools/` | an ops desk with no `bash` and no skills, acting only through four tools declared under `tools:` — release, version bump, service logs, slow requests — covering every parameter type (string, enum, boolean, integer, number), a safe default (`dryRun: true`) and an optional argument; Mellum2 12B on DMR, measured results |
  ## Learn more
  - [Documentation](docs/README.md) — tutorials, how-to guides, reference and explanations, in English and French.
++- [Release notes](release-notes/README.md) — what changed in each version, in English and French.
  - [`guides/`](guides/) — practical guides; for now, the agent settings for small models, in [English](guides/agent-settings-for-small-models.md) and [French](guides/reglages-agents-petits-modeles.md): timeouts, `maxTurns`, `maxOutput`, `agentsMd.maxChars`, context compression, sampling — what each one does, its default, and how to set it for a small local model.
  - The `skills/` directory: Markdown procedures the agent can load with its `read_skill` tool.
  - `agent.lsp.yaml`: the same agent with a language server behind an `lsp` tool (definition, references, hover, symbols, diagnostics, rename) — see [Navigate and refactor code with a language server](docs/en/how-to/navigate-code-with-lsp.md).

added about_tools_test.go +21 -0 Viewed

new file mode 100644
@@ -0,0 +1,21 @@
	1	+package main
	2	+
	3	+import (
	4	+ "strings"
	5	+ "testing"
	6	+
	7	+ "mm/internal/config"
	8	+ "mm/internal/version"
	9	+)
	10	+
	11	+// --about lists the built-in tools by hand; every name the config reserves
	12	+// must be in it, so a new built-in tool cannot be forgotten there again (the
	13	+// list stopped at lsp for seven releases).
	14	+func TestAboutListsEveryBuiltinTool(t *testing.T) {
	15	+ about := version.About()
	16	+ for _, name := range config.BuiltinToolNames {
	17	+ if !strings.Contains(about, name) {
	18	+ t.Errorf("mm --about does not list the built-in tool %q", name)
	19	+ }
	20	+ }
	21	+}

++package main
++
++import (
++	"strings"
++	"testing"
++
++	"mm/internal/config"
++	"mm/internal/version"
++)
++
++// --about lists the built-in tools by hand; every name the config reserves
++// must be in it, so a new built-in tool cannot be forgotten there again (the
++// list stopped at lsp for seven releases).
++func TestAboutListsEveryBuiltinTool(t *testing.T) {
++	about := version.About()
++	for _, name := range config.BuiltinToolNames {
++		if !strings.Contains(about, name) {
++			t.Errorf("mm --about does not list the built-in tool %q", name)
++		}
++	}
++}

added demos/plans/README.en.md +71 -0 Viewed

new file mode 100644
@@ -0,0 +1,71 @@
	1	+# mini-me demo: plans
	2	+
	3	+[Version française](README.fr.md)
	4	+
	5	+Planner builds a tiny Go program step by step — module, code, test, run the tests, README — and works through the steps with its `plan` tool. mm shows where the plan is after every plan action, whatever the model writes in its answer. The model is Qwen3.5 4B, served by [llmman](https://github.com/llmmanorg/llmman). Go must be on the `PATH`.
	6	+
	7	+## What is here
	8	+
	9	+\| File \| Role \|
	10	+\|---\|---\|
	11	+\| `agent.yaml` \| Planner: `planTool: true`, `bash` and the file tools; a system prompt that says how to use the plan (create, then `in_progress` / `done` / `failed` task by task) and to mark `failed`, not fake, a task that needs something missing. \|
	12	+\| `reset.sh` \| Removes `./hello`, the project the agent builds, so the next run starts from nothing. \|
	13	+
	14	+## Run the demo
	15	+
	16	+```bash
	17	+llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
	18	+cd demos/plans
	19	+./reset.sh
	20	+mm agent.yaml
	21	+```
	22	+
	23	+Then ask:
	24	+
	25	+```text
	26	+Do these 4 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. add a function Greet(name string) string and a test for it in main_test.go, 3. run go test and make sure it passes, 4. write a README.md that explains how to run the program.
	27	+```
	28	+
	29	+In the line mode the progress shows between the tool calls (output of a real run):
	30	+
	31	+```text
	32	+📋 plan: create
	33	+📋 [1/4] ⏳ Task 1: Create a Go module named hello with a main.go that prints Hello, mm! (open)
	34	+📋 plan: update
	35	+📋 [2/4] ⏳ Task 2: Add a function Greet(name string) string and a test for it in main_test.go (open)
	36	+📋 plan: update
	37	+📋 [3/4] ⏳ Task 3: Run go test and make sure it passes (open)
	38	+📋 plan: update
	39	+📋 [4/4] ⏳ Task 4: Write a README.md that explains how to run the program (open)
	40	+📋 plan: update
	41	+📋 [4/4] All tasks complete!
	42	+```
	43	+
	44	+In the window (`mm -tui agent.yaml`) each `📋 [C/N] …` line is a notice under the tool call; in an editor the status is inside the tool call's result.
	45	+
	46	+To see a failure, ask for something that cannot be done here:
	47	+
	48	+```text
	49	+Do these 3 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. run go build to check it compiles, 3. push the code to the git remote named origin.
	50	+```
	51	+
	52	+The plan ends on `📋 [3/3] Plan finished: 2 done, 1 failed`.
	53	+
	54	+## Measured results (2026-10-21, `mm -p`, llmman)
	55	+
	56	+Qwen3.5 4B Q4_K_M, `temperature: 0`, thinking off, this `agent.yaml`, a fresh copy of the demo for every run.
	57	+
	58	+\| Request \| Plan shown \| Result \| \|
	59	+\|---\|---\|---\|---\|
	60	+\| the 4 steps above \| `[1/4]` … `[4/4] All tasks complete!` \| `go.mod`, `main.go`, `main_test.go`, `README.md`; `go test` passes (checked after the run) \| 3/3 \|
	61	+\| the 3 steps with a push to `origin` \| `[1/3]` … `[3/3] Plan finished: 2 done, 1 failed` \| module built; the push marked `failed`, the answer says so \| 3/3 \|
	62	+
	63	+64 to 127 s for the 4 steps, 36 to 66 s for the failing case.
	64	+
	65	+## What this demo taught the plan tool
	66	+
	67	+- A small model does not keep the plan's name. In the first runs every `update` came without `name`, failed (`no plan named ''`), and the turn ended in runaway tool calls. Since then an action without a name works on the conversation's latest plan — for every agent, not only this one.
	68	+- Progress is shown by mm. The status line comes from the plan itself, after each action, so it is there even when the model writes nothing about it — and it follows the tasks however the model moves through them (here it marks each task `done` directly, without `in_progress`).
	69	+- "Done" means the model says so. Without the last paragraph of the prompt, asked to push to a remote that does not exist, the model marked the task done 3 times out of 3; in the run inspected, it had created a local bare repository in `/tmp` — outside the project folder — and force-pushed to it. With it, the task is marked `failed` (3/3). It still adds an example remote (`github.com/example/hello.git`) and tries to push before giving up; the push fails without credentials.
	70	+
	71	+See also: [How to manage plans and tasks](../../docs/en/how-to/manage-plans-and-tasks.md), [`plan`](../../docs/en/reference/tools.md#plan).

++# mini-me demo: plans
++
++*[Version française](README.fr.md)*
++
++Planner builds a tiny Go program step by step — module, code, test, run the tests, README — and works through the steps with its `plan` tool. mm shows where the plan is after every plan action, whatever the model writes in its answer. The model is Qwen3.5 4B, served by [llmman](https://github.com/llmmanorg/llmman). Go must be on the `PATH`.
++
++## What is here
++
++| File | Role |
++|---|---|
++| `agent.yaml` | Planner: `planTool: true`, `bash` and the file tools; a system prompt that says how to use the plan (create, then `in_progress` / `done` / `failed` task by task) and to mark `failed`, not fake, a task that needs something missing. |
++| `reset.sh` | Removes `./hello`, the project the agent builds, so the next run starts from nothing. |
++
++## Run the demo
++
++```bash
++llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
++cd demos/plans
++./reset.sh
++mm agent.yaml
++```
++
++Then ask:
++
++```text
++Do these 4 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. add a function Greet(name string) string and a test for it in main_test.go, 3. run go test and make sure it passes, 4. write a README.md that explains how to run the program.
++```
++
++In the line mode the progress shows between the tool calls (output of a real run):
++
++```text
++📋 plan: create
++📋 [1/4] ⏳ Task 1: Create a Go module named hello with a main.go that prints Hello, mm! (open)
++📋 plan: update
++📋 [2/4] ⏳ Task 2: Add a function Greet(name string) string and a test for it in main_test.go (open)
++📋 plan: update
++📋 [3/4] ⏳ Task 3: Run go test and make sure it passes (open)
++📋 plan: update
++📋 [4/4] ⏳ Task 4: Write a README.md that explains how to run the program (open)
++📋 plan: update
++📋 [4/4] All tasks complete!
++```
++
++In the window (`mm -tui agent.yaml`) each `📋 [C/N] …` line is a notice under the tool call; in an editor the status is inside the tool call's result.
++
++To see a failure, ask for something that cannot be done here:
++
++```text
++Do these 3 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. run go build to check it compiles, 3. push the code to the git remote named origin.
++```
++
++The plan ends on `📋 [3/3] Plan finished: 2 done, 1 failed`.
++
++## Measured results (2026-10-21, `mm -p`, llmman)
++
++Qwen3.5 4B Q4_K_M, `temperature: 0`, thinking off, this `agent.yaml`, a fresh copy of the demo for every run.
++
++| Request | Plan shown | Result | |
++|---|---|---|---|
++| the 4 steps above | `[1/4]` … `[4/4] All tasks complete!` | `go.mod`, `main.go`, `main_test.go`, `README.md`; `go test` passes (checked after the run) | 3/3 |
++| the 3 steps with a push to `origin` | `[1/3]` … `[3/3] Plan finished: 2 done, 1 failed` | module built; the push marked `failed`, the answer says so | 3/3 |
++
++64 to 127 s for the 4 steps, 36 to 66 s for the failing case.
++
++## What this demo taught the plan tool
++
++- **A small model does not keep the plan's name.** In the first runs every `update` came without `name`, failed (`no plan named ''`), and the turn ended in runaway tool calls. Since then an action without a name works on the conversation's latest plan — for every agent, not only this one.
++- **Progress is shown by mm.** The status line comes from the plan itself, after each action, so it is there even when the model writes nothing about it — and it follows the tasks however the model moves through them (here it marks each task `done` directly, without `in_progress`).
++- **"Done" means the model says so.** Without the last paragraph of the prompt, asked to push to a remote that does not exist, the model marked the task done 3 times out of 3; in the run inspected, it had created a local bare repository in `/tmp` — outside the project folder — and force-pushed to it. With it, the task is marked `failed` (3/3). It still adds an example remote (`github.com/example/hello.git`) and tries to push before giving up; the push fails without credentials.
++
++See also: [How to manage plans and tasks](../../docs/en/how-to/manage-plans-and-tasks.md), [`plan`](../../docs/en/reference/tools.md#plan).

added demos/plans/README.fr.md +71 -0 Viewed

new file mode 100644
@@ -0,0 +1,71 @@
	1	+# Démo mini-me : les plans
	2	+
	3	+[English version](README.en.md)
	4	+
	5	+Planner construit un petit programme Go étape par étape — module, code, test, lancement des tests, README — et suit les étapes avec son outil `plan`. mm affiche où en est le plan après chaque action sur le plan, quoi que le modèle écrive dans sa réponse. Le modèle est Qwen3.5 4B, servi par [llmman](https://github.com/llmmanorg/llmman). Go doit être dans le `PATH`.
	6	+
	7	+## Contenu
	8	+
	9	+\| Fichier \| Rôle \|
	10	+\|---\|---\|
	11	+\| `agent.yaml` \| Planner : `planTool: true`, `bash` et les outils de fichiers ; un prompt système qui dit comment utiliser le plan (créer, puis `in_progress` / `done` / `failed` tâche par tâche) et de marquer `failed`, sans tricher, une tâche qui demande quelque chose qui manque. \|
	12	+\| `reset.sh` \| Supprime `./hello`, le projet que l'agent construit, pour que l'exécution suivante reparte de zéro. \|
	13	+
	14	+## Lancer la démo
	15	+
	16	+```bash
	17	+llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
	18	+cd demos/plans
	19	+./reset.sh
	20	+mm agent.yaml
	21	+```
	22	+
	23	+Puis demandez :
	24	+
	25	+```text
	26	+Do these 4 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. add a function Greet(name string) string and a test for it in main_test.go, 3. run go test and make sure it passes, 4. write a README.md that explains how to run the program.
	27	+```
	28	+
	29	+En mode ligne, la progression s'affiche entre les appels d'outils (sortie d'une exécution réelle) :
	30	+
	31	+```text
	32	+📋 plan: create
	33	+📋 [1/4] ⏳ Task 1: Create a Go module named hello with a main.go that prints Hello, mm! (open)
	34	+📋 plan: update
	35	+📋 [2/4] ⏳ Task 2: Add a function Greet(name string) string and a test for it in main_test.go (open)
	36	+📋 plan: update
	37	+📋 [3/4] ⏳ Task 3: Run go test and make sure it passes (open)
	38	+📋 plan: update
	39	+📋 [4/4] ⏳ Task 4: Write a README.md that explains how to run the program (open)
	40	+📋 plan: update
	41	+📋 [4/4] All tasks complete!
	42	+```
	43	+
	44	+Dans la fenêtre (`mm -tui agent.yaml`), chaque ligne `📋 [C/N] …` est une notice sous l'appel d'outil ; dans un éditeur, le statut est dans le résultat de l'appel.
	45	+
	46	+Pour voir un échec, demandez quelque chose d'impossible ici :
	47	+
	48	+```text
	49	+Do these 3 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. run go build to check it compiles, 3. push the code to the git remote named origin.
	50	+```
	51	+
	52	+Le plan se termine sur `📋 [3/3] Plan finished: 2 done, 1 failed`.
	53	+
	54	+## Résultats mesurés (2026-10-21, `mm -p`, llmman)
	55	+
	56	+Qwen3.5 4B Q4_K_M, `temperature: 0`, réflexion désactivée, cet `agent.yaml`, une copie neuve de la démo à chaque exécution.
	57	+
	58	+\| Demande \| Plan affiché \| Résultat \| \|
	59	+\|---\|---\|---\|---\|
	60	+\| les 4 étapes ci-dessus \| `[1/4]` … `[4/4] All tasks complete!` \| `go.mod`, `main.go`, `main_test.go`, `README.md` ; `go test` passe (vérifié après l'exécution) \| 3/3 \|
	61	+\| les 3 étapes avec un push vers `origin` \| `[1/3]` … `[3/3] Plan finished: 2 done, 1 failed` \| module construit ; le push marqué `failed`, la réponse le dit \| 3/3 \|
	62	+
	63	+De 64 à 127 s pour les 4 étapes, de 36 à 66 s pour le cas d'échec.
	64	+
	65	+## Ce que cette démo a appris à l'outil plan
	66	+
	67	+- Un petit modèle ne retient pas le nom du plan. Lors des premières exécutions, chaque `update` arrivait sans `name`, échouait (`no plan named ''`), et le tour finissait en appels d'outils emballés. Depuis, une action sans nom porte sur le dernier plan de la conversation — pour tous les agents, pas seulement celui-ci.
	68	+- La progression est affichée par mm. La ligne de statut vient du plan lui-même, après chaque action : elle est là même quand le modèle n'en dit rien — et elle suit les tâches quelle que soit la façon dont le modèle les parcourt (ici il marque chaque tâche `done` directement, sans `in_progress`).
	69	+- « Fait » veut dire « le modèle le dit ». Sans le dernier paragraphe du prompt, invité à pousser vers un dépôt distant qui n'existe pas, le modèle a marqué la tâche faite 3 fois sur 3 ; dans l'exécution examinée, il avait créé un dépôt nu local dans `/tmp` — hors du dossier du projet — et y avait poussé avec `--force`. Avec ce paragraphe, la tâche est marquée `failed` (3/3). Il ajoute encore un remote d'exemple (`github.com/example/hello.git`) et tente de pousser avant d'abandonner ; le push échoue faute d'identifiants.
	70	+
	71	+Voir aussi : [Gérer des plans et des tâches](../../docs/fr/how-to/manage-plans-and-tasks.md), [`plan`](../../docs/fr/reference/tools.md#plan).

++# Démo mini-me : les plans
++
++*[English version](README.en.md)*
++
++Planner construit un petit programme Go étape par étape — module, code, test, lancement des tests, README — et suit les étapes avec son outil `plan`. mm affiche où en est le plan après chaque action sur le plan, quoi que le modèle écrive dans sa réponse. Le modèle est Qwen3.5 4B, servi par [llmman](https://github.com/llmmanorg/llmman). Go doit être dans le `PATH`.
++
++## Contenu
++
++| Fichier | Rôle |
++|---|---|
++| `agent.yaml` | Planner : `planTool: true`, `bash` et les outils de fichiers ; un prompt système qui dit comment utiliser le plan (créer, puis `in_progress` / `done` / `failed` tâche par tâche) et de marquer `failed`, sans tricher, une tâche qui demande quelque chose qui manque. |
++| `reset.sh` | Supprime `./hello`, le projet que l'agent construit, pour que l'exécution suivante reparte de zéro. |
++
++## Lancer la démo
++
++```bash
++llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
++cd demos/plans
++./reset.sh
++mm agent.yaml
++```
++
++Puis demandez :
++
++```text
++Do these 4 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. add a function Greet(name string) string and a test for it in main_test.go, 3. run go test and make sure it passes, 4. write a README.md that explains how to run the program.
++```
++
++En mode ligne, la progression s'affiche entre les appels d'outils (sortie d'une exécution réelle) :
++
++```text
++📋 plan: create
++📋 [1/4] ⏳ Task 1: Create a Go module named hello with a main.go that prints Hello, mm! (open)
++📋 plan: update
++📋 [2/4] ⏳ Task 2: Add a function Greet(name string) string and a test for it in main_test.go (open)
++📋 plan: update
++📋 [3/4] ⏳ Task 3: Run go test and make sure it passes (open)
++📋 plan: update
++📋 [4/4] ⏳ Task 4: Write a README.md that explains how to run the program (open)
++📋 plan: update
++📋 [4/4] All tasks complete!
++```
++
++Dans la fenêtre (`mm -tui agent.yaml`), chaque ligne `📋 [C/N] …` est une notice sous l'appel d'outil ; dans un éditeur, le statut est dans le résultat de l'appel.
++
++Pour voir un échec, demandez quelque chose d'impossible ici :
++
++```text
++Do these 3 things in a new folder ./hello: 1. create a Go module named hello with a main.go that prints Hello, mm!, 2. run go build to check it compiles, 3. push the code to the git remote named origin.
++```
++
++Le plan se termine sur `📋 [3/3] Plan finished: 2 done, 1 failed`.
++
++## Résultats mesurés (2026-10-21, `mm -p`, llmman)
++
++Qwen3.5 4B Q4_K_M, `temperature: 0`, réflexion désactivée, cet `agent.yaml`, une copie neuve de la démo à chaque exécution.
++
++| Demande | Plan affiché | Résultat | |
++|---|---|---|---|
++| les 4 étapes ci-dessus | `[1/4]` … `[4/4] All tasks complete!` | `go.mod`, `main.go`, `main_test.go`, `README.md` ; `go test` passe (vérifié après l'exécution) | 3/3 |
++| les 3 étapes avec un push vers `origin` | `[1/3]` … `[3/3] Plan finished: 2 done, 1 failed` | module construit ; le push marqué `failed`, la réponse le dit | 3/3 |
++
++De 64 à 127 s pour les 4 étapes, de 36 à 66 s pour le cas d'échec.
++
++## Ce que cette démo a appris à l'outil plan
++
++- **Un petit modèle ne retient pas le nom du plan.** Lors des premières exécutions, chaque `update` arrivait sans `name`, échouait (`no plan named ''`), et le tour finissait en appels d'outils emballés. Depuis, une action sans nom porte sur le dernier plan de la conversation — pour tous les agents, pas seulement celui-ci.
++- **La progression est affichée par mm.** La ligne de statut vient du plan lui-même, après chaque action : elle est là même quand le modèle n'en dit rien — et elle suit les tâches quelle que soit la façon dont le modèle les parcourt (ici il marque chaque tâche `done` directement, sans `in_progress`).
++- **« Fait » veut dire « le modèle le dit ».** Sans le dernier paragraphe du prompt, invité à pousser vers un dépôt distant qui n'existe pas, le modèle a marqué la tâche faite 3 fois sur 3 ; dans l'exécution examinée, il avait créé un dépôt nu local dans `/tmp` — hors du dossier du projet — et y avait poussé avec `--force`. Avec ce paragraphe, la tâche est marquée `failed` (3/3). Il ajoute encore un remote d'exemple (`github.com/example/hello.git`) et tente de pousser avant d'abandonner ; le push échoue faute d'identifiants.
++
++Voir aussi : [Gérer des plans et des tâches](../../docs/fr/how-to/manage-plans-and-tasks.md), [`plan`](../../docs/fr/reference/tools.md#plan).

added demos/plans/agent.yaml +73 -0 Viewed

new file mode 100644
@@ -0,0 +1,73 @@
	1	+# Plans — an agent that works through a list of tasks with its `plan` tool,
	2	+# and mm showing where it is after every plan action:
	3	+#
	4	+# 📋 plan: create
	5	+# 📋 [1/4] ⏳ Task 1: Create the Go module … (open)
	6	+# …
	7	+# 📋 [4/4] All tasks complete!
	8	+#
	9	+# The job: build a tiny Go program, step by step, in ./hello — module, code,
	10	+# test, run the tests, README. Needs Go on the PATH.
	11	+#
	12	+# llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
	13	+# llmman serve
	14	+#
	15	+# cd demos/plans
	16	+# ./reset.sh removes ./hello
	17	+# mm -tui agent.yaml or mm -p "…" agent.yaml
	18	+
	19	+provider: llmman
	20	+model: hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
	21	+baseUrl: http://127.0.0.1:17434/v1
	22	+fallback: http://host.docker.internal:17434/v1
	23	+contextWindow: 16384
	24	+
	25	+# The plan tool is on by default; written here because this demo is about it.
	26	+planTool: true
	27	+yolo: true
	28	+bashTool: true
	29	+bashTimeout: 2m
	30	+editTools: true
	31	+askUserTool: false
	32	+skillsDir: no-skills
	33	+
	34	+agentsMd:
	35	+ enabled: false
	36	+
	37	+sampling:
	38	+ temperature: 0.0
	39	+ parallel_tool_calls: false
	40	+ max_tokens: 2048
	41	+ chat_template_kwargs:
	42	+ enable_thinking: false
	43	+
	44	+maxTurns: 30
	45	+watchdogTimeout: 2m
	46	+firstTokenTimeout: 3m
	47	+displayCommands: true
	48	+
	49	+acp:
	50	+ name: planner
	51	+ title: 📋 Planner (plans demo)
	52	+
	53	+system: \|
	54	+ Your name is Planner. You are a coding agent in a terminal, with a bash
	55	+ tool, file tools and a plan tool.
	56	+
	57	+ When the user gives you several tasks, work with a plan:
	58	+ 1. Call plan with action "create" and one task per item of the user's
	59	+ list, in order. Remember the plan name it returns.
	60	+ 2. For each task, in order: call plan "update" with status
	61	+ "in_progress", do the work with your other tools, then call plan
	62	+ "update" with status "done" — or "failed" with a short message if it
	63	+ could not be done, and go on with the next task.
	64	+ 3. When every task is over, answer in two or three sentences: what was
	65	+ done, and what failed if anything.
	66	+
	67	+ Never mark a task done before its work is done. Run commands to check
	68	+ what you did: do not guess their output.
	69	+
	70	+ Work only inside the project folder. If a task needs something that does
	71	+ not exist — a git remote, an account, a server, a file elsewhere — do not
	72	+ create a stand-in to make it pass: mark the task "failed" with a message
	73	+ saying what is missing, and go on.

++# Plans — an agent that works through a list of tasks with its `plan` tool,
++# and mm showing where it is after every plan action:
++#
++#   📋 plan: create
++#   📋 [1/4] ⏳ Task 1: Create the Go module … (open)
++#   …
++#   📋 [4/4] All tasks complete!
++#
++# The job: build a tiny Go program, step by step, in ./hello — module, code,
++# test, run the tests, README. Needs Go on the PATH.
++#
++#   llmman pull hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
++#   llmman serve
++#
++#   cd demos/plans
++#   ./reset.sh                   removes ./hello
++#   mm -tui agent.yaml           or   mm -p "…" agent.yaml
++
++provider: llmman
++model: hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
++baseUrl: http://127.0.0.1:17434/v1
++fallback: http://host.docker.internal:17434/v1
++contextWindow: 16384
++
++# The plan tool is on by default; written here because this demo is about it.
++planTool: true
++yolo: true
++bashTool: true
++bashTimeout: 2m
++editTools: true
++askUserTool: false
++skillsDir: no-skills
++
++agentsMd:
++  enabled: false
++
++sampling:
++  temperature: 0.0
++  parallel_tool_calls: false
++  max_tokens: 2048
++  chat_template_kwargs:
++    enable_thinking: false
++
++maxTurns: 30
++watchdogTimeout: 2m
++firstTokenTimeout: 3m
++displayCommands: true
++
++acp:
++  name: planner
++  title: 📋 Planner (plans demo)
++
++system: |
++  Your name is Planner. You are a coding agent in a terminal, with a bash
++  tool, file tools and a plan tool.
++
++  When the user gives you several tasks, work with a plan:
++  1. Call plan with action "create" and one task per item of the user's
++     list, in order. Remember the plan name it returns.
++  2. For each task, in order: call plan "update" with status
++     "in_progress", do the work with your other tools, then call plan
++     "update" with status "done" — or "failed" with a short message if it
++     could not be done, and go on with the next task.
++  3. When every task is over, answer in two or three sentences: what was
++     done, and what failed if anything.
++
++  Never mark a task done before its work is done. Run commands to check
++  what you did: do not guess their output.
++
++  Work only inside the project folder. If a task needs something that does
++  not exist — a git remote, an account, a server, a file elsewhere — do not
++  create a stand-in to make it pass: mark the task "failed" with a message
++  saying what is missing, and go on.

added demos/plans/reset.sh +7 -0 Viewed

new file mode 100755
@@ -0,0 +1,7 @@
	1	+#!/bin/sh
	2	+# reset.sh — puts the demo back as it ships: removes the ./hello project the
	3	+# agent builds, so the next run starts from nothing.
	4	+set -eu
	5	+cd "$(dirname "$0")"
	6	+rm -rf hello
	7	+echo "reset: ./hello removed"

++#!/bin/sh
++# reset.sh — puts the demo back as it ships: removes the ./hello project the
++# agent builds, so the next run starts from nothing.
++set -eu
++cd "$(dirname "$0")"
++rm -rf hello
++echo "reset: ./hello removed"

modified docs/diagrams/packages.drawio +1 -1 Viewed

@@ -1,3 +1,3 @@
1	1	<mxfile host="app.diagrams.net" modified="2026-10-06T00:00:00.000Z" agent="mm docs" version="24.0.0" type="device"><diagram id="packages" name="packages"><mxGraphModel dx="1400" dy="900" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="1" pageScale="1" pageWidth="1400" pageHeight="1060"><root><mxCell id="0" /><mxCell id="1" parent="0" /><mxCell id="main" value="<b>main (mm)</b><br/><font style="font-size:10px">wiring: config, engine, tools, then one front end</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;" vertex="1" parent="1"><mxGeometry x="420" y="40" width="200" height="60" as="geometry" /></mxCell><mxCell id="agent" value="<b>internal/agent</b><br/><font style="font-size:10px">terminal REPL: /quit /abort /compact /new</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="160" y="180" width="200" height="60" as="geometry" /></mxCell><mxCell id="acp" value="<b>internal/acp</b><br/><font style="font-size:10px">editor front end: Agent Client Protocol on stdio</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="700" y="180" width="200" height="60" as="geometry" /></mxCell><mxCell id="session" value="<b>internal/session</b><br/><font style="font-size:10px">what a conversation starts from; the /new reset</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="compact" value="<b>internal/compact</b><br/><font style="font-size:10px">history compression into a model-written summary</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="40" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="engine" value="<b>internal/engine</b><br/><font style="font-size:10px">LLM connection + generation, one Provider per server</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="700" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="tools" value="<b>internal/tools</b><br/><font style="font-size:10px">built-in tools: bash, read_skill, file tools, lsp, fetch_content; declared toolsbuilt-in tools: bash, read_skill, file tools, lsp, fetch_content</font>lt;/fontbuilt-in tools: bash, read_skill, file tools, lsp, fetch_content</font>gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="detector" value="<b>internal/detector</b><br/><font style="font-size:10px">repeated-action (loop) detection</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="40" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="config" value="<b>internal/config</b><br/><font style="font-size:10px">settings with YAML overrides (agent.yaml)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="spinner" value="<b>internal/spinner</b><br/><font style="font-size:10px">terminal spinner and styling</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="480" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="ui" value="<b>internal/ui</b><br/><font style="font-size:10px">human output destination + shared event sink</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="fileedit" value="<b>internal/fileedit</b><br/><font style="font-size:10px">exact-replacement file editing</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="skills" value="<b>internal/skills</b><br/><font style="font-size:10px">markdown skills catalogue (./skills)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="genkit" value="<b>github.com/firebase/genkit</b><br/><font style="font-size:10px">ai.Message, tools, model calls</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="160" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="acpsdk" value="<b>github.com/coder/acp-go-sdk</b><br/><font style="font-size:10px">JSON-RPC transport for ACP</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="700" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="openai" value="<b>github.com/openai/openai-go</b><br/><font style="font-size:10px">OpenAI-compatible HTTP client</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="yaml" value="<b>gopkg.in/yaml.v3</b><br/><font style="font-size:10px">YAML parsing</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="400" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="e0" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="version" value="<b>internal/version</b><br/><font style="font-size:10px">which mm this is: build stamp, else the commit</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1380" y="600" width="200" height="60" as="geometry" /></mxCell>
2	2	<mxCell id="e61" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="version"><mxGeometry relative="1" as="geometry" /></mxCell>
3		- <mxCell id="e1" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="acp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e2" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e3" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e4" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e5" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e6" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e7" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e8" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e9" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e10" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e11" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e12" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e13" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e14" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e15" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e16" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e17" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e18" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e19" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e20" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="agent" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e21" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="compact" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e22" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="compact" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e23" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="config" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e24" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e25" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e26" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e27" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e28" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="openai"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e29" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e30" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e31" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="fileedit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e32" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e33" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e34" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e35" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="tools" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e36" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="session" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="mention" value="<b>internal/mention</b><br/><font style="font-size:10px">the @path notation of a question</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e37" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e38" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="store" value="<b>internal/store</b><br/><font style="font-size:10px">per-project session files (.mm/sessions/<id>.json), the questions ↑ recalls, /memory and search_memory</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="480" width="200" height="60" as="geometry" /></mxCell><mxCell id="e39" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e40" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="tui" value="<b>internal/tui</b><br/><font style="font-size:10px">window front end: one screen on tcell (mm -tui)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="430" y="160" width="200" height="60" as="geometry" /></mxCell><mxCell id="command" value="<b>internal/command</b><br/><font style="font-size:10px">the one slash-command table, per front end</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="tcell" value="<b>github.com/gdamore/tcell/v2</b><br/><font style="font-size:10px">terminal screen, keys, mouse</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="840" width="200" height="60" as="geometry" /></mxCell><mxCell id="e41" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e42" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e43" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e44" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e45" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e46" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e47" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e48" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e49" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e50" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e51" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tcell"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e52" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e53" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e54" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="command" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="saved" value="<b>internal/saved</b><br/><font style="font-size:10px">an answer to a new file: /save, save_answer</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e87" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e88" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e89" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="clipboard" value="<b>internal/clipboard</b><br/><font style="font-size:10px">system clipboard: native tool, then OSC 52</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e55" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e56" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="lsp" value="<b>internal/lsp</b><br/><font style="font-size:10px">language-server client: definition, references, hover, symbols, diagnostics, rename</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1160" y="600" width="220" height="60" as="geometry" /></mxCell><mxCell id="lspdev" value="<b>go.lsp.dev/protocol + jsonrpc2</b><br/><font style="font-size:10px">LSP types and JSON-RPC 2.0 over stdio</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="1160" y="700" width="220" height="60" as="geometry" /></mxCell><mxCell id="e57" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e58" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e59" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="lsp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e60" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="lsp" target="lspdev"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="legend" value="<b>mm — package dependencies</b><br/>Arrow: "imports". Yellow: entry point. Blue: front ends. Green: shared logic. Grey: leaf packages. Dashed purple: third-party modules.<br/>Regenerate the edge list from: go list -f '{{.ImportPath}}: {{join .Imports " "}}' ./..." style="text;html=1;whiteSpace=wrap;align=left;verticalAlign=top;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="900" y="40" width="440" height="90" as="geometry" /></mxCell><mxCell id="webmd" value="<b>internal/webmd</b><br/><font style="font-size:10px">web page → Markdown: fetch (robots.txt, charset → UTF-8), clean, convert</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="460" width="220" height="60" as="geometry" /></mxCell><mxCell id="xnet" value="<b>golang.org/x/net/html (+ x/text)</b><br/><font style="font-size:10px">HTML5 parser, charset lookup</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="1400" y="760" width="220" height="60" as="geometry" /></mxCell><mxCell id="e_main_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_webmd_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="webmd" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="ask" value="<b>internal/ask</b><br/><font style="font-size:10px">/ask another mm agent: child mm -acp, ACP client side</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="640" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_main_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_command" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_ui" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_acpsdk" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_genkit" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_yaml" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="consent" value="<b>internal/consent</b><br/><font style="font-size:10px">yolo switch, the four answers, per-session memory, the line mode prompt hook</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="500" width="200" height="60" as="geometry" /></mxCell><mxCell id="mcp" value="<b>internal/mcp</b><br/><font style="font-size:10px">MCP clients (stdio, HTTP), their tools wrapped as mm tools</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e62" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e63" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e64" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e65" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e66" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e67" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e68" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e69" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e70" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e71" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e72" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e73" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e74" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="structured" value="<b>internal/structured</b><br/><font style="font-size:10px">structuredOutput: Go declarations → JSON Schema (go/parser)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e75" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="config" target="structured"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="agentsmd" value="<b>internal/agentsmd</b><br/><font style="font-size:10px">the project's AGENTS.md, appended to the system prompt per conversation</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="520.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e76" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e77" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e78" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e79" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e80" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agentsmd" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="elicit" value="<b>internal/elicit</b><br/><font style="font-size:10px">ask_user's questions and answers: normalize, line answers, elicitation form schema</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="780.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e81" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e82" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e83" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e84" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e85" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e86" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="hooks" value="<b>internal/hooks</b><br/><font style="font-size:10px">the hooks: block — runs each hook (sh -c, JSON on stdin/stdout), chain, onError</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="940" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_engine_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_main_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_hooks_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="hooks" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_session" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_engine" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_store" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="docsearch" value="<b>internal/docsearch</b><br/><font style="font-size:10px">BM25 over the text files of folders: /docs, search_docs</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="1640" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="textfold" value="<b>internal/textfold</b><br/><font style="font-size:10px">case and accents folded, rune for rune</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1640" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_docsearch_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_docsearch_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_textfold_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="textfold" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell></root></mxGraphModel></diagram></mxfile>
		\ No newline at end of file
	3	+ <mxCell id="e1" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="acp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e2" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e3" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e4" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e5" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e6" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e7" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e8" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e9" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e10" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e11" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e12" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e13" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e14" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e15" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e16" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e17" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e18" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e19" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e20" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="agent" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e21" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="compact" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e22" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="compact" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e23" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="config" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e24" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e25" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e26" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e27" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e28" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="openai"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e29" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e30" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e31" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="fileedit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e32" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e33" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e34" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e35" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="tools" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e36" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="session" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="mention" value="<b>internal/mention</b><br/><font style="font-size:10px">the @path notation of a question</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e37" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e38" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="store" value="<b>internal/store</b><br/><font style="font-size:10px">per-project session files (.mm/sessions/<id>.json), the questions ↑ recalls, /memory and search_memory</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="480" width="200" height="60" as="geometry" /></mxCell><mxCell id="e39" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e40" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="tui" value="<b>internal/tui</b><br/><font style="font-size:10px">window front end: one screen on tcell (mm -tui)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="430" y="160" width="200" height="60" as="geometry" /></mxCell><mxCell id="command" value="<b>internal/command</b><br/><font style="font-size:10px">the one slash-command table, per front end</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="tcell" value="<b>github.com/gdamore/tcell/v2</b><br/><font style="font-size:10px">terminal screen, keys, mouse</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="840" width="200" height="60" as="geometry" /></mxCell><mxCell id="e41" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e42" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e43" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e44" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e45" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e46" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e47" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e48" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e49" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e50" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e51" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tcell"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e52" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e53" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e54" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="command" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="saved" value="<b>internal/saved</b><br/><font style="font-size:10px">an answer to a new file: /save, save_answer</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e87" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e88" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e89" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="clipboard" value="<b>internal/clipboard</b><br/><font style="font-size:10px">system clipboard: native tool, then OSC 52</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e55" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e56" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="lsp" value="<b>internal/lsp</b><br/><font style="font-size:10px">language-server client: definition, references, hover, symbols, diagnostics, rename</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1160" y="600" width="220" height="60" as="geometry" /></mxCell><mxCell id="lspdev" value="<b>go.lsp.dev/protocol + jsonrpc2</b><br/><font style="font-size:10px">LSP types and JSON-RPC 2.0 over stdio</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="1160" y="700" width="220" height="60" as="geometry" /></mxCell><mxCell id="e57" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e58" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e59" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="lsp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e60" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="lsp" target="lspdev"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="legend" value="<b>mm — package dependencies</b><br/>Arrow: "imports". Yellow: entry point. Blue: front ends. Green: shared logic. Grey: leaf packages. Dashed purple: third-party modules.<br/>Regenerate the edge list from: go list -f '{{.ImportPath}}: {{join .Imports " "}}' ./..." style="text;html=1;whiteSpace=wrap;align=left;verticalAlign=top;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="900" y="40" width="440" height="90" as="geometry" /></mxCell><mxCell id="webmd" value="<b>internal/webmd</b><br/><font style="font-size:10px">web page → Markdown: fetch (robots.txt, charset → UTF-8), clean, convert</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="460" width="220" height="60" as="geometry" /></mxCell><mxCell id="xnet" value="<b>golang.org/x/net/html (+ x/text)</b><br/><font style="font-size:10px">HTML5 parser, charset lookup</font>" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="1400" y="760" width="220" height="60" as="geometry" /></mxCell><mxCell id="e_main_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_webmd_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="webmd" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="ask" value="<b>internal/ask</b><br/><font style="font-size:10px">/ask another mm agent: child mm -acp, ACP client side</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="640" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_main_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_command" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_ui" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_acpsdk" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_genkit" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_yaml" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="consent" value="<b>internal/consent</b><br/><font style="font-size:10px">yolo switch, the four answers, per-session memory, the line mode prompt hook</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="500" width="200" height="60" as="geometry" /></mxCell><mxCell id="mcp" value="<b>internal/mcp</b><br/><font style="font-size:10px">MCP clients (stdio, HTTP), their tools wrapped as mm tools</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e62" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e63" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e64" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e65" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e66" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e67" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e68" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e69" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e70" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e71" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e72" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e73" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e74" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="structured" value="<b>internal/structured</b><br/><font style="font-size:10px">structuredOutput: Go declarations → JSON Schema (go/parser)</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e75" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="config" target="structured"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="agentsmd" value="<b>internal/agentsmd</b><br/><font style="font-size:10px">the project's AGENTS.md, appended to the system prompt per conversation</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="520.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e76" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e77" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e78" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e79" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e80" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agentsmd" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="elicit" value="<b>internal/elicit</b><br/><font style="font-size:10px">ask_user's questions and answers: normalize, line answers, elicitation form schema</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="780.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e81" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e82" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e83" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e84" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e85" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e86" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="hooks" value="<b>internal/hooks</b><br/><font style="font-size:10px">the hooks: block — runs each hook (sh -c, JSON on stdin/stdout), chain, onError</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="940" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_engine_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_main_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_hooks_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="hooks" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_session" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_engine" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_store" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="docsearch" value="<b>internal/docsearch</b><br/><font style="font-size:10px">BM25 over the text files of folders: /docs, search_docs</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="1640" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="textfold" value="<b>internal/textfold</b><br/><font style="font-size:10px">case and accents folded, rune for rune</font>" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1640" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_docsearch_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_docsearch_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_textfold_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="textfold" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_tools" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_tools" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell></root></mxGraphModel></diagram></mxfile>
		\ No newline at end of file

  <mxfile host="app.diagrams.net" modified="2026-10-06T00:00:00.000Z" agent="mm docs" version="24.0.0" type="device"><diagram id="packages" name="packages"><mxGraphModel dx="1400" dy="900" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="1" pageScale="1" pageWidth="1400" pageHeight="1060"><root><mxCell id="0" /><mxCell id="1" parent="0" /><mxCell id="main" value="&lt;b&gt;main (mm)&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;wiring: config, engine, tools, then one front end&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;" vertex="1" parent="1"><mxGeometry x="420" y="40" width="200" height="60" as="geometry" /></mxCell><mxCell id="agent" value="&lt;b&gt;internal/agent&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;terminal REPL: /quit /abort /compact /new&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="160" y="180" width="200" height="60" as="geometry" /></mxCell><mxCell id="acp" value="&lt;b&gt;internal/acp&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;editor front end: Agent Client Protocol on stdio&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="700" y="180" width="200" height="60" as="geometry" /></mxCell><mxCell id="session" value="&lt;b&gt;internal/session&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;what a conversation starts from; the /new reset&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="compact" value="&lt;b&gt;internal/compact&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;history compression into a model-written summary&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="40" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="engine" value="&lt;b&gt;internal/engine&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;LLM connection + generation, one Provider per server&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="700" y="320" width="200" height="60" as="geometry" /></mxCell><mxCell id="tools" value="&lt;b&gt;internal/tools&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;built-in tools: bash, read_skill, file tools, lsp, fetch_content; declared toolsbuilt-in tools: bash, read_skill, file tools, lsp, fetch_content&lt;/font&gt;lt;/fontbuilt-in tools: bash, read_skill, file tools, lsp, fetch_content&lt;/font&gt;gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="detector" value="&lt;b&gt;internal/detector&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;repeated-action (loop) detection&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="40" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="config" value="&lt;b&gt;internal/config&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;settings with YAML overrides (agent.yaml)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="spinner" value="&lt;b&gt;internal/spinner&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;terminal spinner and styling&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="480" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="ui" value="&lt;b&gt;internal/ui&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;human output destination + shared event sink&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="fileedit" value="&lt;b&gt;internal/fileedit&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;exact-replacement file editing&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="skills" value="&lt;b&gt;internal/skills&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;markdown skills catalogue (./skills)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="600" width="200" height="60" as="geometry" /></mxCell><mxCell id="genkit" value="&lt;b&gt;github.com/firebase/genkit&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;ai.Message, tools, model calls&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="160" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="acpsdk" value="&lt;b&gt;github.com/coder/acp-go-sdk&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;JSON-RPC transport for ACP&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="700" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="openai" value="&lt;b&gt;github.com/openai/openai-go&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;OpenAI-compatible HTTP client&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="yaml" value="&lt;b&gt;gopkg.in/yaml.v3&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;YAML parsing&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="400" y="760" width="200" height="60" as="geometry" /></mxCell><mxCell id="e0" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="version" value="&lt;b&gt;internal/version&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;which mm this is: build stamp, else the commit&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1380" y="600" width="200" height="60" as="geometry" /></mxCell>
          <mxCell id="e61" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="version"><mxGeometry relative="1" as="geometry" /></mxCell>
--        <mxCell id="e1" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="acp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e2" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e3" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e4" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e5" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e6" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e7" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e8" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e9" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e10" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e11" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e12" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e13" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e14" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e15" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e16" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e17" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e18" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e19" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e20" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="agent" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e21" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="compact" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e22" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="compact" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e23" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="config" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e24" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e25" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e26" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e27" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e28" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="openai"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e29" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e30" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e31" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="fileedit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e32" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e33" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e34" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e35" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="tools" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e36" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="session" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="mention" value="&lt;b&gt;internal/mention&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the @path notation of a question&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e37" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e38" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="store" value="&lt;b&gt;internal/store&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;per-project session files (.mm/sessions/&lt;id&gt;.json), the questions ↑ recalls, /memory and search_memory&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="480" width="200" height="60" as="geometry" /></mxCell><mxCell id="e39" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e40" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="tui" value="&lt;b&gt;internal/tui&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;window front end: one screen on tcell (mm -tui)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="430" y="160" width="200" height="60" as="geometry" /></mxCell><mxCell id="command" value="&lt;b&gt;internal/command&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the one slash-command table, per front end&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="tcell" value="&lt;b&gt;github.com/gdamore/tcell/v2&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;terminal screen, keys, mouse&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="840" width="200" height="60" as="geometry" /></mxCell><mxCell id="e41" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e42" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e43" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e44" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e45" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e46" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e47" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e48" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e49" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e50" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e51" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tcell"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e52" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e53" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e54" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="command" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="saved" value="&lt;b&gt;internal/saved&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;an answer to a new file: /save, save_answer&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e87" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e88" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e89" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="clipboard" value="&lt;b&gt;internal/clipboard&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;system clipboard: native tool, then OSC 52&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e55" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e56" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="lsp" value="&lt;b&gt;internal/lsp&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;language-server client: definition, references, hover, symbols, diagnostics, rename&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1160" y="600" width="220" height="60" as="geometry" /></mxCell><mxCell id="lspdev" value="&lt;b&gt;go.lsp.dev/protocol + jsonrpc2&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;LSP types and JSON-RPC 2.0 over stdio&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="1160" y="700" width="220" height="60" as="geometry" /></mxCell><mxCell id="e57" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e58" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e59" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="lsp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e60" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="lsp" target="lspdev"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="legend" value="&lt;b&gt;mm — package dependencies&lt;/b&gt;&lt;br/&gt;Arrow: &quot;imports&quot;. Yellow: entry point. Blue: front ends. Green: shared logic. Grey: leaf packages. Dashed purple: third-party modules.&lt;br/&gt;Regenerate the edge list from: go list -f '{{.ImportPath}}: {{join .Imports &quot; &quot;}}' ./..." style="text;html=1;whiteSpace=wrap;align=left;verticalAlign=top;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="900" y="40" width="440" height="90" as="geometry" /></mxCell><mxCell id="webmd" value="&lt;b&gt;internal/webmd&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;web page → Markdown: fetch (robots.txt, charset → UTF-8), clean, convert&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="460" width="220" height="60" as="geometry" /></mxCell><mxCell id="xnet" value="&lt;b&gt;golang.org/x/net/html (+ x/text)&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;HTML5 parser, charset lookup&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="1400" y="760" width="220" height="60" as="geometry" /></mxCell><mxCell id="e_main_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_webmd_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="webmd" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="ask" value="&lt;b&gt;internal/ask&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;/ask another mm agent: child mm -acp, ACP client side&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="640" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_main_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_command" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_ui" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_acpsdk" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_genkit" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_yaml" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="consent" value="&lt;b&gt;internal/consent&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;yolo switch, the four answers, per-session memory, the line mode prompt hook&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="500" width="200" height="60" as="geometry" /></mxCell><mxCell id="mcp" value="&lt;b&gt;internal/mcp&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;MCP clients (stdio, HTTP), their tools wrapped as mm tools&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e62" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e63" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e64" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e65" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e66" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e67" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e68" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e69" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e70" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e71" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e72" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e73" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e74" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="structured" value="&lt;b&gt;internal/structured&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;structuredOutput: Go declarations → JSON Schema (go/parser)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e75" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="config" target="structured"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="agentsmd" value="&lt;b&gt;internal/agentsmd&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the project's AGENTS.md, appended to the system prompt per conversation&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="520.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e76" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e77" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e78" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e79" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e80" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agentsmd" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="elicit" value="&lt;b&gt;internal/elicit&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;ask_user's questions and answers: normalize, line answers, elicitation form schema&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="780.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e81" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e82" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e83" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e84" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e85" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e86" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="hooks" value="&lt;b&gt;internal/hooks&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the hooks: block — runs each hook (sh -c, JSON on stdin/stdout), chain, onError&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="940" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_engine_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_main_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_hooks_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="hooks" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_session" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_engine" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_store" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="docsearch" value="&lt;b&gt;internal/docsearch&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;BM25 over the text files of folders: /docs, search_docs&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="1640" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="textfold" value="&lt;b&gt;internal/textfold&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;case and accents folded, rune for rune&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1640" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_docsearch_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_docsearch_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_textfold_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="textfold" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell></root></mxGraphModel></diagram></mxfile>
 \ No newline at end of file
++        <mxCell id="e1" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="acp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e2" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e3" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e4" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e5" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e6" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e7" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e8" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e9" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e10" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e11" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e12" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e13" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="acp" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e14" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e15" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e16" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e17" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e18" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e19" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e20" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="agent" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e21" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="compact" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e22" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="compact" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e23" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="config" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e24" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e25" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e26" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e27" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e28" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="engine" target="openai"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e29" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e30" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e31" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="fileedit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e32" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="skills"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e33" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="spinner"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e34" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e35" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="tools" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e36" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="session" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="mention" value="&lt;b&gt;internal/mention&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the @path notation of a question&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e37" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e38" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="store" value="&lt;b&gt;internal/store&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;per-project session files (.mm/sessions/&lt;id&gt;.json), the questions ↑ recalls, /memory and search_memory&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="480" width="200" height="60" as="geometry" /></mxCell><mxCell id="e39" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e40" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="tui" value="&lt;b&gt;internal/tui&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;window front end: one screen on tcell (mm -tui)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1"><mxGeometry x="430" y="160" width="200" height="60" as="geometry" /></mxCell><mxCell id="command" value="&lt;b&gt;internal/command&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the one slash-command table, per front end&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="tcell" value="&lt;b&gt;github.com/gdamore/tcell/v2&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;terminal screen, keys, mouse&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="920" y="840" width="200" height="60" as="geometry" /></mxCell><mxCell id="e41" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="tui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e42" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e43" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e44" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e45" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e46" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="mention"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e47" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e48" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e49" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e50" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e51" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tcell"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e52" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e53" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e54" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="command" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="saved" value="&lt;b&gt;internal/saved&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;an answer to a new file: /save, save_answer&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="920" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e87" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e88" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e89" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="saved"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="clipboard" value="&lt;b&gt;internal/clipboard&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;system clipboard: native tool, then OSC 52&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1140" y="680" width="200" height="60" as="geometry" /></mxCell><mxCell id="e55" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e56" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="clipboard"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="lsp" value="&lt;b&gt;internal/lsp&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;language-server client: definition, references, hover, symbols, diagnostics, rename&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1160" y="600" width="220" height="60" as="geometry" /></mxCell><mxCell id="lspdev" value="&lt;b&gt;go.lsp.dev/protocol + jsonrpc2&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;LSP types and JSON-RPC 2.0 over stdio&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="1160" y="700" width="220" height="60" as="geometry" /></mxCell><mxCell id="e57" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e58" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="lsp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e59" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="lsp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e60" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="lsp" target="lspdev"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="legend" value="&lt;b&gt;mm — package dependencies&lt;/b&gt;&lt;br/&gt;Arrow: &quot;imports&quot;. Yellow: entry point. Blue: front ends. Green: shared logic. Grey: leaf packages. Dashed purple: third-party modules.&lt;br/&gt;Regenerate the edge list from: go list -f '{{.ImportPath}}: {{join .Imports &quot; &quot;}}' ./..." style="text;html=1;whiteSpace=wrap;align=left;verticalAlign=top;fillColor=#ffffff;strokeColor=#999999;" vertex="1" parent="1"><mxGeometry x="900" y="40" width="440" height="90" as="geometry" /></mxCell><mxCell id="webmd" value="&lt;b&gt;internal/webmd&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;web page → Markdown: fetch (robots.txt, charset → UTF-8), clean, convert&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="460" width="220" height="60" as="geometry" /></mxCell><mxCell id="xnet" value="&lt;b&gt;golang.org/x/net/html (+ x/text)&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;HTML5 parser, charset lookup&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"><mxGeometry x="1400" y="760" width="220" height="60" as="geometry" /></mxCell><mxCell id="e_main_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_webmd" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="webmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_webmd_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="webmd" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="ask" value="&lt;b&gt;internal/ask&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;/ask another mm agent: child mm -acp, ACP client side&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="430" y="640" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_main_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_ask" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="ask"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_command" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="command"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_ui" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_acpsdk" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="acpsdk"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_genkit" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="genkit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_yaml" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="ask" target="yaml"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="consent" value="&lt;b&gt;internal/consent&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;yolo switch, the four answers, per-session memory, the line mode prompt hook&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="700" y="500" width="200" height="60" as="geometry" /></mxCell><mxCell id="mcp" value="&lt;b&gt;internal/mcp&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;MCP clients (stdio, HTTP), their tools wrapped as mm tools&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1400" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e62" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e63" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="consent" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e64" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e65" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e66" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e67" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e68" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e69" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="consent"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e70" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e71" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="ui"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e72" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="mcp" target="detector"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e73" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e74" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="mcp"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="structured" value="&lt;b&gt;internal/structured&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;structuredOutput: Go declarations → JSON Schema (go/parser)&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="260" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e75" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="config" target="structured"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="agentsmd" value="&lt;b&gt;internal/agentsmd&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the project's AGENTS.md, appended to the system prompt per conversation&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="520.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e76" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e77" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e78" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e79" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="agentsmd"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e80" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agentsmd" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="elicit" value="&lt;b&gt;internal/elicit&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;ask_user's questions and answers: normalize, line answers, elicitation form schema&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="780.0" y="940.0" width="200" height="60" as="geometry" /></mxCell><mxCell id="e81" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e82" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e83" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e84" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e85" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e86" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="elicit"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="hooks" value="&lt;b&gt;internal/hooks&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;the hooks: block — runs each hook (sh -c, JSON on stdin/stdout), chain, onError&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="940" y="400" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_engine_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="engine" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_ask_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="ask" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_main_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="main" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_hooks_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="hooks" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_session" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="session"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_compact" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="compact"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_engine" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="engine"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_hooks" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="hooks"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_store" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="store"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="docsearch" value="&lt;b&gt;internal/docsearch&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;BM25 over the text files of folders: /docs, search_docs&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"><mxGeometry x="1640" y="460" width="200" height="60" as="geometry" /></mxCell><mxCell id="textfold" value="&lt;b&gt;internal/textfold&lt;/b&gt;&lt;br/&gt;&lt;font style=&quot;font-size:10px&quot;&gt;case and accents folded, rune for rune&lt;/font&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;strokeColor=#666666;" vertex="1" parent="1"><mxGeometry x="1640" y="560" width="200" height="60" as="geometry" /></mxCell><mxCell id="e_docsearch_config" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="config"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_docsearch_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="docsearch" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_store_textfold" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="store" target="textfold"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_textfold_xnet" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;dashed=1;" edge="1" parent="1" source="textfold" target="xnet"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tools_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tools" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_agent_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="agent" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_docsearch" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="docsearch"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_tui_tools" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="tui" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell><mxCell id="e_acp_tools" style="endArrow=block;html=1;rounded=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="1" source="acp" target="tools"><mxGeometry relative="1" as="geometry" /></mxCell></root></mxGraphModel></diagram></mxfile>
 \ No newline at end of file

modified docs/en/README.md +2 -1 Viewed

@@ -26,6 +26,7 @@ This documentation is organised along the four [Diátaxis](https://diataxis.fr)
26	26	- [Run mini-me in a Docker Sandbox](how-to/run-in-a-docker-sandbox.md) — the `mini-me` kit from [mini-me-kit](https://rickub.com/bots-garden/mini-me-kit), the `go-toolchain` and `cli-tools` mixins under `sbx/`, window mode and editor
27	27	- [Add a skill](how-to/add-a-skill.md)
28	28	- [Declare a custom tool](how-to/declare-a-custom-tool.md)
	29	+- [Manage plans and tasks](how-to/manage-plans-and-tasks.md) — the `plan` tool: give a list, follow the `📋` progress line, plans per conversation
29	30	- [Save an answer to a file](how-to/save-an-answer.md)
30	31	- [Search a documentation folder](how-to/search-the-docs.md) — `/docs <words>`, the `search_docs` tool, `docSearch` (BM25, no model)
31	32	- [Search earlier sessions](how-to/search-earlier-sessions.md) — `/memory <words>`, the `search_memory` tool, `sessions.searchMemory` / `sessions.autoSearchMemory`
@@ -54,7 +55,7 @@ This documentation is organised along the four [Diátaxis](https://diataxis.fr)
54	55	- [Command line](reference/cli.md) — flags, environment variables, start-up banner
55	56	- [Slash commands](reference/slash-commands.md) — `/quit`, `/abort`, `/compact`, `/new`, `/sessions`, `/resume`, `/help`, keyboard shortcuts, window keys
56	57	- [Configuration file](reference/configuration.md) — every YAML key, its default and its validation rule
57		-- [Built-in tools](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, declared tools
	58	+- [Built-in tools](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `plan`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, `search_docs`, `search_memory`, declared tools
58	59	- [Skill file format](reference/skill-format.md)
59	60	- [Hooks](reference/hooks.md) — the six events, the JSON a hook reads and writes, `matcher`, `timeout`, `onError`, validation errors
60	61	- [Agent Client Protocol surface](reference/acp.md)

  - [Run mini-me in a Docker Sandbox](how-to/run-in-a-docker-sandbox.md) — the `mini-me` kit from [mini-me-kit](https://rickub.com/bots-garden/mini-me-kit), the `go-toolchain` and `cli-tools` mixins under `sbx/`, window mode and editor
  - [Add a skill](how-to/add-a-skill.md)
  - [Declare a custom tool](how-to/declare-a-custom-tool.md)
++- [Manage plans and tasks](how-to/manage-plans-and-tasks.md) — the `plan` tool: give a list, follow the `📋` progress line, plans per conversation
  - [Save an answer to a file](how-to/save-an-answer.md)
  - [Search a documentation folder](how-to/search-the-docs.md) — `/docs <words>`, the `search_docs` tool, `docSearch` (BM25, no model)
  - [Search earlier sessions](how-to/search-earlier-sessions.md) — `/memory <words>`, the `search_memory` tool, `sessions.searchMemory` / `sessions.autoSearchMemory`
  - [Command line](reference/cli.md) — flags, environment variables, start-up banner
  - [Slash commands](reference/slash-commands.md) — `/quit`, `/abort`, `/compact`, `/new`, `/sessions`, `/resume`, `/help`, keyboard shortcuts, window keys
  - [Configuration file](reference/configuration.md) — every YAML key, its default and its validation rule
--- [Built-in tools](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, declared tools
++- [Built-in tools](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `plan`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, `search_docs`, `search_memory`, declared tools
  - [Skill file format](reference/skill-format.md)
  - [Hooks](reference/hooks.md) — the six events, the JSON a hook reads and writes, `matcher`, `timeout`, `onError`, validation errors
  - [Agent Client Protocol surface](reference/acp.md)

modified docs/en/how-to/manage-plans-and-tasks.md +42 -427 Viewed

@@ -1,450 +1,65 @@
1		----
2		-title: Manage Plans and Tasks
3		----
	1	+# How to manage plans and tasks
4	2
5		-# Manage Plans and Tasks
	3	+This guide shows how to have the agent work through a list of tasks with its `plan` tool, and how to follow its progress. The tool is on by default (`planTool: true`).
6	4
7		-Mini-me includes a built-in plan tool that lets the agent create, track, and manage multi-step todo lists. When you give the agent a list of tasks to do, it automatically creates a plan and tracks progress, showing you the current status after each response.
	5	+## Give the agent a list
8	6
9		-## Quick Example
	7	+Ask for several things at once, as a list:
10	8
	9	+```text
	10	+Do these 2 things: 1. create a file a.txt containing A, 2. create a file b.txt containing B.
11	11	```
12		-You: Do these 3 things: 1. Create a README file, 2. Add a license, 3. Commit the changes
13	12
14		-Agent: I'll handle this in 3 steps. Starting with: Create a README file
15		-📋 plan: create user-request-1
	13	+The model creates a plan from the list, works through it, and updates each task's status as it goes. Small models do it more reliably when you say so: "Use the plan tool. Do these…".
16	14
17		-Agent: README file created.
18		-📋 [2/3] 🔄 Task 2: Add a license (in_progress)
	15	+## Follow the progress
19	16
20		-Agent: License added.
21		-📋 [3/3] 🔄 Task 3: Commit the changes (in_progress)
	17	+mm shows where the plan is after every plan action that changes or reads its state (`create`, `update`, `next`, `status`) — whether or not the model mentions it in its answer.
22	18
23		-Agent: All changes committed.
24		-📋 [3/3] ✅ All tasks complete!
25		-```
26		-
27		-The `📋` line shows you exactly where the agent is in your plan.
28		-
29		----
30		-
31		-## How It Works
32		-
33		-### The Plan Lifecycle
34		-
35		-1. You state your request as a list of tasks
36		-2. Agent detects the plan and creates it using the `plan` tool
37		-3. Agent works on tasks one by one, updating status
38		-4. Agent reports progress after each action
39		-5. Plan completes when all tasks are done
40		-
41		-### Status Icons
42		-
43		-\| Icon \| Meaning \|
44		-\|------\|---------\|
45		-\| ⏳ \| Task not started (open) \|
46		-\| 🔄 \| Task currently in progress \|
47		-\| ✅ \| Task completed successfully \|
48		-\| ❌ \| Task failed \|
49		-
50		-### Status Line Format
51		-
52		-```
53		-📋 [current/total] icon Task N: description (status)
54		-```
55		-
56		-Examples:
57		-- `📋 [1/5] ⏳ Task 1: Analyze the code (open)` - First task, not started yet
58		-- `📋 [2/5] 🔄 Task 2: Fix the bugs (in_progress)` - Second task, currently working
59		-- `📋 [5/5] ✅ All tasks complete!` - All done!
60		-- `📋 [3/5] ❌ Task 3: Run tests (failed: connection timeout)` - Task failed with error
61		-
62		----
63		-
64		-## User Scenarios
65		-
66		-### Scenario 1: Numbered List
67		-
68		-Your input:
69		-```
70		-Do these 5 things:
71		-1. Check the current version
72		-2. Update dependencies
73		-3. Run tests
74		-4. Fix any failures
75		-5. Deploy to staging
76		-```
77		-
78		-What happens:
79		-- Agent creates a plan with 5 tasks
80		-- Works through each sequentially
81		-- Shows progress: `📋 [1/5]`, `📋 [2/5]`, etc.
82		-- Completes with: `📋 [5/5] ✅ All tasks complete!`
83		-
84		----
85		-
86		-### Scenario 2: Comma-Separated Tasks
87		-
88		-Your input:
89		-```
90		-Set up the project: create the directory, initialize git, and add a README
91		-```
92		-
93		-What happens:
94		-- Agent detects this as a multi-step request
95		-- Creates a plan with 3 tasks
96		-- Executes in order
97		-
98		----
99		-
100		-### Scenario 3: Multiple Plans
101		-
102		-You can have multiple active plans in a single session.
103		-
104		-Your input:
105		-```
106		-First, fix the auth module: 1. Find the bug, 2. Write the fix, 3. Test it
107		-```
108		-
109		-Agent works on the auth plan, shows: `📋 [2/3] 🔄 Task 2: Write the fix (in_progress)`
110		-
111		-Then you say:
112		-```
113		-Also, update the documentation: 1. Update README, 2. Update CHANGELOG
114		-```
115		-
116		-Agent creates a second plan and manages both independently.
117		-
118		-To see all active plans:
119		-```
120		-What plans do I have?
121		-```
122		-
123		----
124		-
125		-## Interacting with Plans
126		-
127		-### Check Current Plan Status
128		-
129		-Ask: "What's the status of my plan?"
130		-
131		-Agent responds:
132		-```
133		-Current plan: [2/3] 🔄 Task 2: Write the fix (in_progress). Next: Test it
134		-```
135		-
136		-### Skip a Task
137		-
138		-Ask: "Skip the test step, just deploy"
139		-
140		-Agent:
141		-- Marks the current task as done
142		-- Skips to the next task
143		-- Confirms: "Skipping test step. Deploying to staging..."
144		-
145		-### Abandon a Plan
146		-
147		-Ask: "Never mind, I don't need this anymore"
148		-
149		-Agent:
150		-- Asks for confirmation
151		-- If you confirm, deletes the plan
152		-- All progress is lost
153		-
154		----
155		-
156		-## Plan Management Commands
157		-
158		-While the plan tool is primarily model-driven (the agent uses it automatically), you can also interact with plans directly:
159		-
160		-### List All Plans
161		-
162		-Command: `/plans` (not yet implemented in v1 - coming in v2)
163		-
164		-What it shows:
165		-```
166		-Active plans:
167		-1. setup-project (2/5 tasks complete)
168		-2. fix-bugs (0/3 tasks complete)
169		-```
170		-
171		-### View Plan Details
172		-
173		-Ask the agent: "Show me the details of plan X"
174		-
175		-Agent responds: Full task list with status for each task
176		-
177		-### Delete a Plan
178		-
179		-Ask the agent: "Delete plan X"
180		-
181		-Agent: Confirms and deletes the plan
182		-
183		----
184		-
185		-## Session Lifecycle
186		-
187		-### Plans and /new Command
188		-
189		-When you start a new session with `/new`:
190		-- All active plans are cleared
191		-- This is intentional: a new session means a fresh start
192		-- If you want to continue an old plan, don't use `/new`
193		-
194		-### Plans and Agent Restart
195		-
196		-In v1 (current version):
197		-- Plans are stored in memory only
198		-- Plans are lost when the agent restarts
199		-- This keeps v1 simple and fast
200		-
201		-In v2 (future):
202		-- Plans will be persisted to session files
203		-- Plans will survive agent restarts
204		-
205		----
206		-
207		-## The plan Tool (For Advanced Users)
208		-
209		-The `plan` tool is available for the model to call. It supports these actions:
210		-
211		-\| Action \| Purpose \| Required Parameters \|
212		-\|--------\|---------\|---------------------\|
213		-\| `create` \| Create a new plan \| `tasks` (array of `{description}`) \|
214		-\| `get` \| Get plan details \| `name` \|
215		-\| `update` \| Update task status \| `name`, `taskId`, `status` \|
216		-\| `next` \| Move to next task \| `name` \|
217		-\| `list` \| List all plans \| none \|
218		-\| `delete` \| Delete a plan \| `name` \|
219		-\| `status` \| Get status line \| `name` \|
	19	+- Line mode (`mm agent.yaml`, `mm -p`): one line before each call, one line after it. Measured on Qwen3.5 4B with the request above:
220	20
221		-### Status Values
	21	+ ```text
	22	+ 📋 plan: create
	23	+ 📋 [1/2] ⏳ Task 1: Create a file a.txt containing A (open)
	24	+ 💾 write_file: a.txt (1 bytes)
	25	+ 💾 write_file: b.txt (1 bytes)
	26	+ 📋 plan: update user-request-1
	27	+ 📋 [2/2] ⏳ Task 2: Create a file b.txt containing B (open)
	28	+ 📋 plan: update user-request-1
	29	+ 📋 [2/2] All tasks complete!
	30	+ ```
222	31
223		-- `open` - Task not started
224		-- `in_progress` - Task currently being worked on
225		-- `done` - Task completed successfully
226		-- `failed` - Task failed
	32	+- Window (`mm -tui`): each call is a tool call in the conversation, and the `📋 [C/N] …` line follows it as a notice.
	33	+- Editor (`mm -acp`): each call is a tool call; the status line is inside its result.
227	34
228		----
	35	+The status line shows the task being worked on — the first one in progress, else the first one still open: `[C/N] <icon> Task C: <description> (<status>)`. Icons: ⏳ open, 🔄 in progress, ✅ done, ❌ failed. When every task is over it says `[N/N] All tasks complete!`, or `[N/N] Plan finished: D done, F failed` if some failed.
229	36
230		-## The plan-manager Skill
	37	+## Steer the plan
231	38
232		-The agent uses the `plan-manager` skill to know when and how to use the plan tool. This skill includes:
	39	+The model decides when to call the tool; you steer it in your own words:
233	40
234		-- Detection patterns: How to recognize when a user gives a multi-step request
235		-- Usage guidelines: When to create plans, update tasks, move between tasks
236		-- Display format: How to show plan status to the user
237		-- Error handling: What to do when things go wrong
	41	+- Where are we? "What's the status of the plan?" — the model calls `status` and mm shows the line.
	42	+- Skip a task. "Skip the tests and go on." — the model can mark it and move on with `next`.
	43	+- A task failed. The model can mark it `failed`, with a note; the status line then moves to the next open task.
	44	+- Drop the plan. "Forget this plan." — the model can call `delete`.
238	45
239		-You don't need to manually load this skill - the agent does it automatically when appropriate.
	46	+There is no slash command for plans: they are the model's tool.
240	47
241		----
	48	+## Variants
242	49
243		-## Examples
244		-
245		-### Example 1: Bug Fixing Workflow
246		-
247		-```
248		-You: Fix these issues in the auth module:
249		- 1. Check for null pointer exceptions
250		- 2. Add input validation
251		- 3. Write unit tests
252		- 4. Update the documentation
253		-
254		-Agent: I'll handle this in 4 steps. Starting with: Check for null pointer exceptions
255		-📋 plan: create user-request-1
256		-
257		-... (agent checks code) ...
258		-
259		-Agent: Found 3 potential null pointer issues.
260		-📋 [2/4] 🔄 Task 2: Add input validation (in_progress)
261		-
262		-... (agent adds validation) ...
263		-
264		-Agent: Input validation added to auth.go.
265		-📋 [3/4] 🔄 Task 3: Write unit tests (in_progress)
266		-
267		-... (agent writes tests) ...
268		-
269		-Agent: Tests written and passing.
270		-📋 [4/4] 🔄 Task 4: Update the documentation (in_progress)
271		-
272		-... (agent updates docs) ...
273		-
274		-Agent: Documentation updated with new validation requirements.
275		-📋 [4/4] ✅ All tasks complete!
276		-```
277		-
278		----
279		-
280		-### Example 2: Project Setup
281		-
282		-```
283		-You: Set up a new Go project with:
284		- 1. Module initialization
285		- 2. Directory structure
286		- 3. Main file
287		- 4. Test file
288		-
289		-Agent: I'll set this up in 4 steps. Starting with: Module initialization
290		-📋 plan: create user-request-1
291		-
292		-Agent: Module initialized with go mod
293		-📋 [2/4] 🔄 Task 2: Directory structure (in_progress)
294		-
295		-Agent: Directories created (cmd, internal, pkg, tests)
296		-📋 [3/4] 🔄 Task 3: Main file (in_progress)
297		-
298		-Agent: Main file created with basic structure
299		-📋 [4/4] 🔄 Task 4: Test file (in_progress)
300		-
301		-Agent: Test file created with sample tests
302		-📋 [4/4] ✅ All tasks complete!
303		-```
304		-
305		----
306		-
307		-### Example 3: Code Review
308		-
309		-```
310		-You: Review this PR and:
311		- 1. Check for security issues
312		- 2. Verify tests pass
313		- 3. Review the documentation
314		- 4. Leave comments
315		-
316		-Agent: I'll review this PR in 4 steps. Starting with: Check for security issues
317		-📋 plan: create user-request-1
318		-
319		-Agent: No security issues found.
320		-📋 [2/4] 🔄 Task 2: Verify tests pass (in_progress)
321		-
322		-Agent: All tests passing.
323		-📋 [3/4] 🔄 Task 3: Review the documentation (in_progress)
324		-
325		-Agent: Documentation looks good, just needs a typo fix.
326		-📋 [4/4] 🔄 Task 4: Leave comments (in_progress)
327		-
328		-Agent: Comments left on the PR.
329		-📋 [4/4] ✅ All tasks complete!
330		-```
331		-
332		----
333		-
334		-### Example 4: With a Task Failure
335		-
336		-```
337		-You: Do these:
338		- 1. Connect to the database
339		- 2. Run the migration
340		- 3. Import the data
341		-
342		-Agent: I'll handle this in 3 steps. Starting with: Connect to the database
343		-📋 plan: create user-request-1
344		-
345		-Agent: Database connection established.
346		-📋 [2/3] 🔄 Task 2: Run the migration (in_progress)
347		-
348		-Agent: Migration completed.
349		-📋 [3/3] 🔄 Task 3: Import the data (in_progress)
350		-
351		-Agent: Error: connection timeout to data source
352		-📋 plan: update user-request-1
353		-
354		-Agent: ⚠️ Task 3 failed: connection timeout to data source
355		-📋 [3/3] ❌ Task 3: Import the data (failed: connection timeout)
356		-
357		-Agent: 2 of 3 tasks completed. The data import failed due to a connection timeout. Would you like me to retry, or skip this task?
358		-```
359		-
360		----
361		-
362		-## Configuration
363		-
364		-The plan tool can be enabled or disabled via the YAML configuration file:
365		-
366		-```yaml
367		-# In your agent.yaml file
368		-planTool: true # Enabled by default
369		-```
370		-
371		-To disable:
372		-```yaml
373		-planTool: false
374		-```
375		-
376		-When disabled, the agent will not have access to the `plan` tool and will not automatically create or track plans.
377		-
378		----
	50	+- Several conversations in an editor. Each thread has its own plans: a plan made in one Zed thread is not visible in another.
	51	+- Starting over. `/new` and `/resume` forget the plans in every front end; an editor's reload (`session/load`, `session/resume`) starts the thread with no plan. Plans are kept in memory only, never in the session file, and are lost when mm stops.
	52	+- The `plan-manager` skill. `skills/plan-manager` tells the model when to make a plan and how to move through it. Put it in the agent's `skillsDir` for a model that does not use the tool by itself. It also asks the model to repeat the status line in its answer; mm shows it anyway.
	53	+- An agent that does not need plans. Set `planTool: false`: the tool is not declared. Do it for an agent whose job is a single answer (writing a document, answering questions), where a small model may otherwise loop on plan calls.
379	54
380	55	## Troubleshooting
381	56
382		-### Agent Not Creating Plans
383		-
384		-If the agent isn't creating plans for your multi-step requests:
385		-
386		-1. Make your request clearer - Use numbered lists
387		- - ❌ "Fix some things"
388		- - ✅ "Do these: 1. Fix A, 2. Fix B, 3. Fix C"
389		-
390		-2. Be more explicit - Use words like "plan", "steps", "then"
391		- - ✅ "My plan is to first do X, then do Y"
392		-
393		-3. Check the skill is loaded - The agent should automatically load the `plan-manager` skill when it detects a plan-like request
394		-
395		-### The Agent Keeps Calling `plan update`
396		-
397		-`update` changes the status of one task; it cannot change a plan's tasks. A model that sends an `update` without a valid `status` — often with a `tasks` list, trying to edit the plan — gets an answer that quotes its call back, names the missing field, says that `tasks` is only read by `create`, and gives a valid call:
398		-
399		-```
400		-[error: plan update changes the status of ONE task, and "status" is missing. You sent: {…}. "status" must be one of: open, in_progress, done, failed. "tasks" is ignored by update (only create reads it): to change a plan's tasks, delete the plan and create it again. A valid call: {"action":"update","name":"user-request-1","taskId":1,"status":"done"}]
401		-```
402		-
403		-If a small model still repeats the call, set `planTool: false` for an agent whose job does not need plans (writing a document, answering questions): the model then works without them.
404		-
405		-### Plans Disappearing
406		-
407		-In v1, plans are stored in memory and will be lost when:
408		-- You start a new session with `/new`
409		-- The agent restarts
410		-- You close and reopen the terminal
411		-
412		-This is expected behavior in v1. Session persistence is coming in v2.
413		-
414		----
415		-
416		-## Best Practices
417		-
418		-1. Be specific - The more detail in your task descriptions, the better the agent can execute them
419		-
420		-2. Keep plans focused - One plan per logical request. Don't mix unrelated tasks
421		-
422		-3. Check progress - The status line shows you exactly where the agent is
423		-
424		-4. Ask for clarification - If the agent seems stuck or confused, ask "What's the status?"
425		-
426		-5. Start small - For complex multi-step work, start with a few tasks and add more as needed
427		-
428		----
429		-
430		-## Comparison with Other Tools
431		-
432		-\| Feature \| plan tool \| skills \| slash commands \|
433		-\|---------\|-----------\|--------\|----------------\|
434		-\| Purpose \| Track multi-step work \| Load procedures \| User commands \|
435		-\| Trigger \| Automatic (model detects) \| Explicit (`/skill`) \| Explicit (`/command`) \|
436		-\| Persistence (v1) \| In-memory, session only \| File-based \| N/A \|
437		-\| User interaction \| Indirect (via model) \| Indirect (model loads) \| Direct \|
438		-
439		-The `plan` tool complements other mini-me features:
440		-- Use skills for complex, reusable procedures
441		-- Use plan tool for ad-hoc multi-step requests
442		-- Use slash commands for direct user actions
443		-
444		----
	57	+- The agent does not create a plan. Give a numbered list, and say "use the plan tool". With a small model, add the `plan-manager` skill.
	58	+- The agent keeps calling `plan update`. `update` changes the status of ONE task; it cannot change a plan's tasks. A call without a valid `status` gets an answer that quotes the call back, names the missing field and gives a valid call. If a small model still repeats it, set `planTool: false` for that agent.
445	59
446		-## See Also
	60	+## See also
447	61
448		-- [Slash Commands Reference](../reference/slash-commands.md) - For user commands
449		-- [Tools Reference](../reference/tools.md) - For all built-in tools
450		-- [Skills Overview](../explanation/skills.md) - For the skill system
	62	+- [Tools reference — `plan`](../reference/tools.md#plan): every action, field, result and error
	63	+- [Configuration reference](../reference/configuration.md): `planTool`
	64	+- [Slash commands reference](../reference/slash-commands.md): `/new`, `/resume`
	65	+- `demos/plans/` — a measured demo: a Go program built in four planned steps, and a plan that ends with a failed task

-----
++# How to manage plans and tasks
--title: Manage Plans and Tasks
-----
--# Manage Plans and Tasks
++This guide shows how to have the agent work through a list of tasks with its `plan` tool, and how to follow its progress. The tool is on by default (`planTool: true`).
--Mini-me includes a built-in **plan** tool that lets the agent create, track, and manage multi-step todo lists. When you give the agent a list of tasks to do, it automatically creates a plan and tracks progress, showing you the current status after each response.
++## Give the agent a list
--## Quick Example
++Ask for several things at once, as a list:
++```text
++Do these 2 things: 1. create a file a.txt containing A, 2. create a file b.txt containing B.
  ```
--You: Do these 3 things: 1. Create a README file, 2. Add a license, 3. Commit the changes
--Agent: I'll handle this in 3 steps. Starting with: Create a README file
++The model creates a plan from the list, works through it, and updates each task's status as it goes. Small models do it more reliably when you say so: "Use the plan tool. Do these…".
--📋 plan: create user-request-1
--Agent: README file created.
++## Follow the progress
--📋 [2/3] 🔄 Task 2: Add a license (in_progress)
--Agent: License added.
++mm shows where the plan is after every plan action that changes or reads its state (`create`, `update`, `next`, `status`) — whether or not the model mentions it in its answer.
--📋 [3/3] 🔄 Task 3: Commit the changes (in_progress)
--Agent: All changes committed.
++- **Line mode** (`mm agent.yaml`, `mm -p`): one line before each call, one line after it. Measured on Qwen3.5 4B with the request above:
--📋 [3/3] ✅ All tasks complete!
--```
--
--The `📋` line shows you exactly where the agent is in your plan.
--
-----
--
--## How It Works
--
--### The Plan Lifecycle
--
--1. **You state your request** as a list of tasks
--2. **Agent detects the plan** and creates it using the `plan` tool
--3. **Agent works on tasks** one by one, updating status
--4. **Agent reports progress** after each action
--5. **Plan completes** when all tasks are done
--
--### Status Icons
--
--| Icon | Meaning |
--|------|---------|
--| ⏳ | Task not started (open) |
--| 🔄 | Task currently in progress |
--| ✅ | Task completed successfully |
--| ❌ | Task failed |
--
--### Status Line Format
--
--```
--📋 [current/total] icon Task N: description (status)
--```
--
--Examples:
--- `📋 [1/5] ⏳ Task 1: Analyze the code (open)` - First task, not started yet
--- `📋 [2/5] 🔄 Task 2: Fix the bugs (in_progress)` - Second task, currently working
--- `📋 [5/5] ✅ All tasks complete!` - All done!
--- `📋 [3/5] ❌ Task 3: Run tests (failed: connection timeout)` - Task failed with error
--
-----
--
--## User Scenarios
--
--### Scenario 1: Numbered List
--
--**Your input:**
--```
--Do these 5 things:
--1. Check the current version
--2. Update dependencies
--3. Run tests
--4. Fix any failures
--5. Deploy to staging
--```
--
--**What happens:**
--- Agent creates a plan with 5 tasks
--- Works through each sequentially
--- Shows progress: `📋 [1/5]`, `📋 [2/5]`, etc.
--- Completes with: `📋 [5/5] ✅ All tasks complete!`
--
-----
--
--### Scenario 2: Comma-Separated Tasks
--
--**Your input:**
--```
--Set up the project: create the directory, initialize git, and add a README
--```
--
--**What happens:**
--- Agent detects this as a multi-step request
--- Creates a plan with 3 tasks
--- Executes in order
--
-----
--
--### Scenario 3: Multiple Plans
--
--You can have multiple active plans in a single session.
--
--**Your input:**
--```
--First, fix the auth module: 1. Find the bug, 2. Write the fix, 3. Test it
--```
--
--Agent works on the auth plan, shows: `📋 [2/3] 🔄 Task 2: Write the fix (in_progress)`
--
--**Then you say:**
--```
--Also, update the documentation: 1. Update README, 2. Update CHANGELOG
--```
--
--Agent creates a second plan and manages both independently.
--
--**To see all active plans:**
--```
--What plans do I have?
--```
--
-----
--
--## Interacting with Plans
--
--### Check Current Plan Status
--
--**Ask:** "What's the status of my plan?"
--
--**Agent responds:**
--```
--Current plan: [2/3] 🔄 Task 2: Write the fix (in_progress). Next: Test it
--```
--
--### Skip a Task
--
--**Ask:** "Skip the test step, just deploy"
--
--**Agent:**
--- Marks the current task as done
--- Skips to the next task
--- Confirms: "Skipping test step. Deploying to staging..."
--
--### Abandon a Plan
--
--**Ask:** "Never mind, I don't need this anymore"
--
--**Agent:**
--- Asks for confirmation
--- If you confirm, deletes the plan
--- All progress is lost
--
-----
--
--## Plan Management Commands
--
--While the plan tool is primarily model-driven (the agent uses it automatically), you can also interact with plans directly:
--
--### List All Plans
--
--**Command:** `/plans` (not yet implemented in v1 - coming in v2)
--
--**What it shows:**
--```
--Active plans:
--1. setup-project (2/5 tasks complete)
--2. fix-bugs (0/3 tasks complete)
--```
--
--### View Plan Details
--
--**Ask the agent:** "Show me the details of plan X"
--
--**Agent responds:** Full task list with status for each task
--
--### Delete a Plan
--
--**Ask the agent:** "Delete plan X"
--
--**Agent:** Confirms and deletes the plan
--
-----
--
--## Session Lifecycle
--
--### Plans and /new Command
--
--When you start a new session with `/new`:
--- All active plans are **cleared**
--- This is intentional: a new session means a fresh start
--- If you want to continue an old plan, don't use `/new`
--
--### Plans and Agent Restart
--
--In v1 (current version):
--- Plans are stored **in memory only**
--- Plans are **lost** when the agent restarts
--- This keeps v1 simple and fast
--
--In v2 (future):
--- Plans will be persisted to session files
--- Plans will survive agent restarts
--
-----
--
--## The plan Tool (For Advanced Users)
--
--The `plan` tool is available for the model to call. It supports these actions:
--
--| Action | Purpose | Required Parameters |
--|--------|---------|---------------------|
--| `create` | Create a new plan | `tasks` (array of `{description}`) |
--| `get` | Get plan details | `name` |
--| `update` | Update task status | `name`, `taskId`, `status` |
--| `next` | Move to next task | `name` |
--| `list` | List all plans | none |
--| `delete` | Delete a plan | `name` |
--| `status` | Get status line | `name` |
--### Status Values
++  ```text
++  📋 plan: create
++  📋 [1/2] ⏳ Task 1: Create a file a.txt containing A (open)
++  💾 write_file: a.txt (1 bytes)
++  💾 write_file: b.txt (1 bytes)
++  📋 plan: update user-request-1
++  📋 [2/2] ⏳ Task 2: Create a file b.txt containing B (open)
++  📋 plan: update user-request-1
++  📋 [2/2] All tasks complete!
++  ```
--- `open` - Task not started
++- **Window** (`mm -tui`): each call is a tool call in the conversation, and the `📋 [C/N] …` line follows it as a notice.
--- `in_progress` - Task currently being worked on
++- **Editor** (`mm -acp`): each call is a tool call; the status line is inside its result.
--- `done` - Task completed successfully
--- `failed` - Task failed
-----
++The status line shows the task being worked on — the first one in progress, else the first one still open: `[C/N] <icon> Task C: <description> (<status>)`. Icons: ⏳ open, 🔄 in progress, ✅ done, ❌ failed. When every task is over it says `[N/N] All tasks complete!`, or `[N/N] Plan finished: D done, F failed` if some failed.
--## The plan-manager Skill
++## Steer the plan
--The agent uses the `plan-manager` skill to know **when and how** to use the plan tool. This skill includes:
++The model decides when to call the tool; you steer it in your own words:
--- **Detection patterns:** How to recognize when a user gives a multi-step request
++- **Where are we?** "What's the status of the plan?" — the model calls `status` and mm shows the line.
--- **Usage guidelines:** When to create plans, update tasks, move between tasks
++- **Skip a task.** "Skip the tests and go on." — the model can mark it and move on with `next`.
--- **Display format:** How to show plan status to the user
++- **A task failed.** The model can mark it `failed`, with a note; the status line then moves to the next open task.
--- **Error handling:** What to do when things go wrong
++- **Drop the plan.** "Forget this plan." — the model can call `delete`.
--You don't need to manually load this skill - the agent does it automatically when appropriate.
++There is no slash command for plans: they are the model's tool.
-----
++## Variants
--## Examples
++- **Several conversations in an editor.** Each thread has its own plans: a plan made in one Zed thread is not visible in another.
--
++- **Starting over.** `/new` and `/resume` forget the plans in every front end; an editor's reload (`session/load`, `session/resume`) starts the thread with no plan. Plans are kept in memory only, never in the session file, and are lost when mm stops.
--### Example 1: Bug Fixing Workflow
++- **The `plan-manager` skill.** `skills/plan-manager` tells the model when to make a plan and how to move through it. Put it in the agent's `skillsDir` for a model that does not use the tool by itself. It also asks the model to repeat the status line in its answer; mm shows it anyway.
--
++- **An agent that does not need plans.** Set `planTool: false`: the tool is not declared. Do it for an agent whose job is a single answer (writing a document, answering questions), where a small model may otherwise loop on plan calls.
--```
--You: Fix these issues in the auth module:
--     1. Check for null pointer exceptions
--     2. Add input validation
--     3. Write unit tests
--     4. Update the documentation
--
--Agent: I'll handle this in 4 steps. Starting with: Check for null pointer exceptions
--📋 plan: create user-request-1
--
--... (agent checks code) ...
--
--Agent: Found 3 potential null pointer issues.
--📋 [2/4] 🔄 Task 2: Add input validation (in_progress)
--
--... (agent adds validation) ...
--
--Agent: Input validation added to auth.go.
--📋 [3/4] 🔄 Task 3: Write unit tests (in_progress)
--
--... (agent writes tests) ...
--
--Agent: Tests written and passing.
--📋 [4/4] 🔄 Task 4: Update the documentation (in_progress)
--
--... (agent updates docs) ...
--
--Agent: Documentation updated with new validation requirements.
--📋 [4/4] ✅ All tasks complete!
--```
--
-----
--
--### Example 2: Project Setup
--
--```
--You: Set up a new Go project with:
--     1. Module initialization
--     2. Directory structure
--     3. Main file
--     4. Test file
--
--Agent: I'll set this up in 4 steps. Starting with: Module initialization
--📋 plan: create user-request-1
--
--Agent: Module initialized with go mod
--📋 [2/4] 🔄 Task 2: Directory structure (in_progress)
--
--Agent: Directories created (cmd, internal, pkg, tests)
--📋 [3/4] 🔄 Task 3: Main file (in_progress)
--
--Agent: Main file created with basic structure
--📋 [4/4] 🔄 Task 4: Test file (in_progress)
--
--Agent: Test file created with sample tests
--📋 [4/4] ✅ All tasks complete!
--```
--
-----
--
--### Example 3: Code Review
--
--```
--You: Review this PR and:
--     1. Check for security issues
--     2. Verify tests pass
--     3. Review the documentation
--     4. Leave comments
--
--Agent: I'll review this PR in 4 steps. Starting with: Check for security issues
--📋 plan: create user-request-1
--
--Agent: No security issues found.
--📋 [2/4] 🔄 Task 2: Verify tests pass (in_progress)
--
--Agent: All tests passing.
--📋 [3/4] 🔄 Task 3: Review the documentation (in_progress)
--
--Agent: Documentation looks good, just needs a typo fix.
--📋 [4/4] 🔄 Task 4: Leave comments (in_progress)
--
--Agent: Comments left on the PR.
--📋 [4/4] ✅ All tasks complete!
--```
--
-----
--
--### Example 4: With a Task Failure
--
--```
--You: Do these:
--     1. Connect to the database
--     2. Run the migration
--     3. Import the data
--
--Agent: I'll handle this in 3 steps. Starting with: Connect to the database
--📋 plan: create user-request-1
--
--Agent: Database connection established.
--📋 [2/3] 🔄 Task 2: Run the migration (in_progress)
--
--Agent: Migration completed.
--📋 [3/3] 🔄 Task 3: Import the data (in_progress)
--
--Agent: Error: connection timeout to data source
--📋 plan: update user-request-1
--
--Agent: ⚠️ Task 3 failed: connection timeout to data source
--📋 [3/3] ❌ Task 3: Import the data (failed: connection timeout)
--
--Agent: 2 of 3 tasks completed. The data import failed due to a connection timeout. Would you like me to retry, or skip this task?
--```
--
-----
--
--## Configuration
--
--The plan tool can be enabled or disabled via the YAML configuration file:
--
--```yaml
--# In your agent.yaml file
--planTool: true  # Enabled by default
--```
--
--To disable:
--```yaml
--planTool: false
--```
--
--When disabled, the agent will not have access to the `plan` tool and will not automatically create or track plans.
--
-----
  ## Troubleshooting
--### Agent Not Creating Plans
++- **The agent does not create a plan.** Give a numbered list, and say "use the plan tool". With a small model, add the `plan-manager` skill.
--
++- **The agent keeps calling `plan update`.** `update` changes the status of ONE task; it cannot change a plan's tasks. A call without a valid `status` gets an answer that quotes the call back, names the missing field and gives a valid call. If a small model still repeats it, set `planTool: false` for that agent.
--If the agent isn't creating plans for your multi-step requests:
--
--1. **Make your request clearer** - Use numbered lists
--   - ❌ "Fix some things"
--   - ✅ "Do these: 1. Fix A, 2. Fix B, 3. Fix C"
--
--2. **Be more explicit** - Use words like "plan", "steps", "then"
--   - ✅ "My plan is to first do X, then do Y"
--
--3. **Check the skill is loaded** - The agent should automatically load the `plan-manager` skill when it detects a plan-like request
--
--### The Agent Keeps Calling `plan update`
--
--`update` changes the status of one task; it cannot change a plan's tasks. A model that sends an `update` without a valid `status` — often with a `tasks` list, trying to edit the plan — gets an answer that quotes its call back, names the missing field, says that `tasks` is only read by `create`, and gives a valid call:
--
--```
--[error: plan update changes the status of ONE task, and "status" is missing. You sent: {…}. "status" must be one of: open, in_progress, done, failed. "tasks" is ignored by update (only create reads it): to change a plan's tasks, delete the plan and create it again. A valid call: {"action":"update","name":"user-request-1","taskId":1,"status":"done"}]
--```
--
--If a small model still repeats the call, set `planTool: false` for an agent whose job does not need plans (writing a document, answering questions): the model then works without them.
--
--### Plans Disappearing
--
--In v1, plans are stored in memory and will be lost when:
--- You start a new session with `/new`
--- The agent restarts
--- You close and reopen the terminal
--
--This is expected behavior in v1. Session persistence is coming in v2.
--
-----
--
--## Best Practices
--
--1. **Be specific** - The more detail in your task descriptions, the better the agent can execute them
--
--2. **Keep plans focused** - One plan per logical request. Don't mix unrelated tasks
--
--3. **Check progress** - The status line shows you exactly where the agent is
--
--4. **Ask for clarification** - If the agent seems stuck or confused, ask "What's the status?"
--
--5. **Start small** - For complex multi-step work, start with a few tasks and add more as needed
--
-----
--
--## Comparison with Other Tools
--
--| Feature | plan tool | skills | slash commands |
--|---------|-----------|--------|----------------|
--| Purpose | Track multi-step work | Load procedures | User commands |
--| Trigger | Automatic (model detects) | Explicit (`/skill`) | Explicit (`/command`) |
--| Persistence (v1) | In-memory, session only | File-based | N/A |
--| User interaction | Indirect (via model) | Indirect (model loads) | Direct |
--
--The `plan` tool complements other mini-me features:
--- Use **skills** for complex, reusable procedures
--- Use **plan tool** for ad-hoc multi-step requests
--- Use **slash commands** for direct user actions
--
-----
--## See Also
++## See also
--- [Slash Commands Reference](../reference/slash-commands.md) - For user commands
++- [Tools reference — `plan`](../reference/tools.md#plan): every action, field, result and error
--- [Tools Reference](../reference/tools.md) - For all built-in tools
++- [Configuration reference](../reference/configuration.md): `planTool`
--- [Skills Overview](../explanation/skills.md) - For the skill system
++- [Slash commands reference](../reference/slash-commands.md): `/new`, `/resume`
++- `demos/plans/` — a measured demo: a Go program built in four planned steps, and a plan that ends with a failed task

modified docs/en/reference/slash-commands.md +3 -3 Viewed

@@ -9,9 +9,9 @@
9	9	\| `/quit` \| Exits the program. \| Cancels the generation, prints `Goodbye.`, exits. \|
10	10	\| `/abort` \| Prints `⚠️ No generation in progress.` \| Cancels the generation, prints `🛑 Aborting...`; the commands already run stay in the history. \|
11	11	\| `/compact` \| Compresses the history now, regardless of the threshold; prints the `🗜️` report, or `🗜️ nothing to compact` when no turn is older than `context.keepLastTurns`. \| Sent to the model as a new message (see below). \|
12		-\| `/new` \| Clears the history back to the system prompt, resets the loop detector and forgets the server's last token count; prints `🆕 new session: N message(s) forgotten` (N counts every message except the system prompt). \| Treated like any other text: cancels the generation, keeps its partial history, then runs as `/new` on that history. \|
	12	+\| `/new` \| Clears the history back to the system prompt, resets the loop detector, forgets the server's last token count and the conversation's plans; prints `🆕 new session: N message(s) forgotten` (N counts every message except the system prompt). \| Treated like any other text: cancels the generation, keeps its partial history, then runs as `/new` on that history. \|
13	13	\| `/sessions` \| Lists the project's stored sessions, newest first: id, last update, question count, first question, then ` · ⚠️ N error(s)` when the session recorded errors; `*` marks the current one. With `sessions.enabled: false`: `⚠️ sessions are off (set sessions.enabled: true in the YAML)`. \| Sent to the model as a new message. \|
14		-\| `/resume [id]` \| Continues a stored session: the most recently updated one other than the current, skipping empty ones, or the named id. Prints `↩️ resumed <id> — N message(s), started with "<first question>"` (plus ` · ⚠️ N error(s)` when the session recorded errors); resets the loop detector and the server's token count; later turns are saved to that session. Refused with a `⚠️` reason when sessions are off, when there is no earlier session, when the id is unknown or is the current session. \| Sent to the model as a new message. \|
	14	+\| `/resume [id]` \| Continues a stored session: the most recently updated one other than the current, skipping empty ones, or the named id. Prints `↩️ resumed <id> — N message(s), started with "<first question>"` (plus ` · ⚠️ N error(s)` when the session recorded errors); resets the loop detector, the server's token count and the plans; later turns are saved to that session. Refused with a `⚠️` reason when sessions are off, when there is no earlier session, when the id is unknown or is the current session. \| Sent to the model as a new message. \|
15	15	\| `/help` \| Prints the commands of this front end, one per line with their description, aligned. \| Sent to the model as a new message. \|
16	16	\| `/copy [n\\|all]` \| Copies the last answer (`n`: the n-th from the end) to the clipboard through `pbcopy`/`wl-copy`/`xclip`/`xsel` or, failing that, OSC 52 when stdout is a terminal, and prints `📋 copied N chars (<road>)` or `⚠️ …`. `/copy all` is window-mode only: here it says so. \| Sent to the model as a new message. \|
17	17	\| `/save <file> [n]` \| Writes the last answer (`n`: the n-th from the end) as it is to a new file, creating the missing directories, and prints `💾 saved the last answer to <file> (N bytes)`. Never overwrites: an existing file gives `⚠️ <file> already exists — nothing was written; choose another name`. A file name with spaces is not supported. The model is not called. \| Sent to the model as a new message. \|
@@ -176,7 +176,7 @@ Two commands exist in ACP mode, three with `unloadModels` (the default): `/new`,
176	176
177	177	\| Command \| Over ACP \|
178	178	\|---------\|----------\|
179		-\| `/new` \| Same as in the terminal: history back to the system prompt, token count forgotten; the session id, `cwd` and allow-always grants are kept. \|
	179	+\| `/new` \| Same as in the terminal: history back to the system prompt, token count and this session's plans forgotten (other sessions keep theirs); the session id, `cwd` and allow-always grants are kept. \|
180	180	\| `/compact` \| Same as in the terminal, always forced: `🗜️ compressed …`, `🗜️ nothing to compact: …` or `[compact: failed, history kept: …]`. Waits for a running turn to finish first. \|
181	181	\| `/memory <words>` \| With `sessions.enabled` and `sessions.searchMemory`, advertised with the input hint `<words>`: same as in the terminal, in the session's project (`sessions.dir` under its `cwd`), as the agent's text; the history is not touched. \|
182	182	\| `/docs <words>` \| With `docSearch.searchDocs`, advertised with the input hint `<words>`: same as in the terminal, from the session's `cwd`, as the agent's text; the history is not touched. \|

  | `/quit` | Exits the program. | Cancels the generation, prints `Goodbye.`, exits. |
  | `/abort` | Prints `⚠️ No generation in progress.` | Cancels the generation, prints `🛑 Aborting...`; the commands already run stay in the history. |
  | `/compact` | Compresses the history now, regardless of the threshold; prints the `🗜️` report, or `🗜️ nothing to compact` when no turn is older than `context.keepLastTurns`. | Sent to the model as a new message (see below). |
--| `/new` | Clears the history back to the system prompt, resets the loop detector and forgets the server's last token count; prints `🆕 new session: N message(s) forgotten` (N counts every message except the system prompt). | Treated like any other text: cancels the generation, keeps its partial history, then runs as `/new` on that history. |
++| `/new` | Clears the history back to the system prompt, resets the loop detector, forgets the server's last token count and the conversation's plans; prints `🆕 new session: N message(s) forgotten` (N counts every message except the system prompt). | Treated like any other text: cancels the generation, keeps its partial history, then runs as `/new` on that history. |
  | `/sessions` | Lists the project's stored sessions, newest first: id, last update, question count, first question, then ` · ⚠️ N error(s)` when the session recorded errors; `*` marks the current one. With `sessions.enabled: false`: `⚠️ sessions are off (set sessions.enabled: true in the YAML)`. | Sent to the model as a new message. |
--| `/resume [id]` | Continues a stored session: the most recently updated one other than the current, skipping empty ones, or the named id. Prints `↩️ resumed <id> — N message(s), started with "<first question>"` (plus ` · ⚠️ N error(s)` when the session recorded errors); resets the loop detector and the server's token count; later turns are saved to that session. Refused with a `⚠️` reason when sessions are off, when there is no earlier session, when the id is unknown or is the current session. | Sent to the model as a new message. |
++| `/resume [id]` | Continues a stored session: the most recently updated one other than the current, skipping empty ones, or the named id. Prints `↩️ resumed <id> — N message(s), started with "<first question>"` (plus ` · ⚠️ N error(s)` when the session recorded errors); resets the loop detector, the server's token count and the plans; later turns are saved to that session. Refused with a `⚠️` reason when sessions are off, when there is no earlier session, when the id is unknown or is the current session. | Sent to the model as a new message. |
  | `/help` | Prints the commands of this front end, one per line with their description, aligned. | Sent to the model as a new message. |
  | `/copy [n\|all]` | Copies the last answer (`n`: the n-th from the end) to the clipboard through `pbcopy`/`wl-copy`/`xclip`/`xsel` or, failing that, OSC 52 when stdout is a terminal, and prints `📋 copied N chars (<road>)` or `⚠️ …`. `/copy all` is window-mode only: here it says so. | Sent to the model as a new message. |
  | `/save <file> [n]` | Writes the last answer (`n`: the n-th from the end) as it is to a new file, creating the missing directories, and prints `💾 saved the last answer to <file> (N bytes)`. Never overwrites: an existing file gives `⚠️ <file> already exists — nothing was written; choose another name`. A file name with spaces is not supported. The model is not called. | Sent to the model as a new message. |
  | Command | Over ACP |
  |---------|----------|
--| `/new` | Same as in the terminal: history back to the system prompt, token count forgotten; the session id, `cwd` and allow-always grants are kept. |
++| `/new` | Same as in the terminal: history back to the system prompt, token count and this session's plans forgotten (other sessions keep theirs); the session id, `cwd` and allow-always grants are kept. |
  | `/compact` | Same as in the terminal, always forced: `🗜️ compressed …`, `🗜️ nothing to compact: …` or `[compact: failed, history kept: …]`. Waits for a running turn to finish first. |
  | `/memory <words>` | With `sessions.enabled` and `sessions.searchMemory`, advertised with the input hint `<words>`: same as in the terminal, in the session's project (`sessions.dir` under its `cwd`), as the agent's text; the history is not touched. |
  | `/docs <words>` | With `docSearch.searchDocs`, advertised with the input hint `<words>`: same as in the terminal, from the session's `cwd`, as the agent's text; the history is not touched. |

modified docs/en/reference/tools.md +32 -0 Viewed

@@ -132,6 +132,38 @@ the closest the file comes to it — copy these lines:
132	132	│ 6 │ }
133	133	```
134	134
	135	+## `plan`
	136	+
	137	+Declared when `planTool: true` (the default). Turns a list of tasks into a plan the model updates as it works. See [Manage plans and tasks](../how-to/manage-plans-and-tasks.md).
	138	+
	139	+\| Field \| Type \| Description \|
	140	+\|-------\|------\|-------------\|
	141	+\| `action` \| string \| Required. `create`, `get`, `update`, `next`, `list`, `delete` or `status`. \|
	142	+\| `name` \| string \| The plan. For `create`, optional: a plan without a name is called `user-request-N` (N counts the conversation's plans). For the other actions, optional too: without it, the action works on the conversation's most recently created plan (`list` ignores it). \|
	143	+\| `tasks` \| list of `{description, id?}` \| For `create`: the tasks, in order. A task without an `id` gets its position (1, 2, …). Ignored by the other actions. \|
	144	+\| `taskId` \| integer \| For `update`: the task, from 1. \|
	145	+\| `status` \| string \| For `update`: `open`, `in_progress`, `done` or `failed`. \|
	146	+\| `message` \| string \| For `update`: a note kept on the task. \|
	147	+
	148	+\| Action \| Does \| Result (JSON text) \|
	149	+\|--------\|------\|--------------------\|
	150	+\| `create` \| Creates the plan, every task `open`. Refused when the name is taken. \| `result` (`Plan '<name>' created with N tasks. Current: Task 1`), `planName`, `status`, `current`, `total`; in the line mode, also the whole `plan`. \|
	151	+\| `get` \| Reads one plan. \| `result`, `plan` (name, tasks with id, description, status, start and end times, note). \|
	152	+\| `update` \| Sets one task's status; `in_progress` records its start time (once) and makes it the current task, `done` and `failed` record its end time. \| `result`, `status` (the status line, below). \|
	153	+\| `next` \| Marks the current task `done` and starts the next one that is neither done nor failed (`in_progress`); a task already finished with `update` is skipped. \| `result` (`Task N done. Next: Task M - <description>`, or `Plan '<name>' is complete! All N tasks done.`), `status`, `nextTask` when there is one. \|
	154	+\| `list` \| Lists the plans held. \| `result`, `plans` (names, oldest first). \|
	155	+\| `delete` \| Removes a plan. \| `result`. \|
	156	+\| `status` \| Reads the status line. \| `result` (`Plan '<name>': D/N tasks complete`), `status`. \|
	157	+
	158	+\| Aspect \| Behaviour \|
	159	+\|--------\|-----------\|
	160	+\| Status line \| `[C/N] <icon> Task C: <description> (<status>)` for the task being worked on — the first one in progress, else the first one still open —, `[N/N] Plan finished: D done, F failed` when every task is over and some failed, `[N/N] All tasks complete!` when all are done. Icons: ⏳ open, 🔄 in progress, ✅ done, ❌ failed. \|
	161	+\| Errors \| Returned as text, never as a failed generation: `[error: no plans found for this session]`, `[error: no plan named '<name>' found]`, `[error: plan '<name>' already exists]`, `[error: no task with id N in plan '<name>']`, `[error: unknown action '<action>'. Valid actions: …]`. An `update` without a valid `status` quotes the call back, says that `tasks` is ignored by `update` when it was sent, and gives a call that works. \|
	162	+\| Storage \| In memory, per conversation: the line mode and the window hold one; over ACP each session has its own, and one thread never sees another's plans. Plans are not written to the session file and are lost when mm stops. A conversation that changes starts with no plan: `/new` and `/resume` in every front end, `session/load` and `session/resume` over ACP. \|
	163	+\| Permission \| `create`, `update`, `next` and `delete` are requested when `yolo: false`; a refusal returns `[plan: command rejected by the user]`. `get`, `list` and `status` never ask. \|
	164	+\| Display \| Line mode: `📋 plan: <action> <name>` before the call, then `📋 <status line>` after every action that has one (`create`, `update`, `next`, `status`). Window: the call as a tool call, and `📋 <status line>` as a notice. Editor (ACP): the call as a tool call, the status inside its result. The progress is shown by mm, whether or not the model repeats it. \|
	165	+\| ACP and window \| Tool call of kind `edit` for `create`, `update`, `next`, `delete`, `read` for the others; a result starting with `[error:` is a failed call. \|
	166	+
135	167	## `lsp`
136	168
137	169	Declared when `lsp.servers` lists at least one server whose command is on the `PATH` at start-up. A listed server that is not found is skipped with a `[warning: lsp server <name> (<command>) not found on the PATH — skipped]`; when none is found the tool is not declared and the banner says so.

    │ 6 │ }
  ```
++## `plan`
++
++Declared when `planTool: true` (the default). Turns a list of tasks into a plan the model updates as it works. See [Manage plans and tasks](../how-to/manage-plans-and-tasks.md).
++
++| Field | Type | Description |
++|-------|------|-------------|
++| `action` | string | Required. `create`, `get`, `update`, `next`, `list`, `delete` or `status`. |
++| `name` | string | The plan. For `create`, optional: a plan without a name is called `user-request-N` (N counts the conversation's plans). For the other actions, optional too: without it, the action works on the conversation's most recently created plan (`list` ignores it). |
++| `tasks` | list of `{description, id?}` | For `create`: the tasks, in order. A task without an `id` gets its position (1, 2, …). Ignored by the other actions. |
++| `taskId` | integer | For `update`: the task, from 1. |
++| `status` | string | For `update`: `open`, `in_progress`, `done` or `failed`. |
++| `message` | string | For `update`: a note kept on the task. |
++
++| Action | Does | Result (JSON text) |
++|--------|------|--------------------|
++| `create` | Creates the plan, every task `open`. Refused when the name is taken. | `result` (`Plan '<name>' created with N tasks. Current: Task 1`), `planName`, `status`, `current`, `total`; in the line mode, also the whole `plan`. |
++| `get` | Reads one plan. | `result`, `plan` (name, tasks with id, description, status, start and end times, note). |
++| `update` | Sets one task's status; `in_progress` records its start time (once) and makes it the current task, `done` and `failed` record its end time. | `result`, `status` (the status line, below). |
++| `next` | Marks the current task `done` and starts the next one that is neither done nor failed (`in_progress`); a task already finished with `update` is skipped. | `result` (`Task N done. Next: Task M - <description>`, or `Plan '<name>' is complete! All N tasks done.`), `status`, `nextTask` when there is one. |
++| `list` | Lists the plans held. | `result`, `plans` (names, oldest first). |
++| `delete` | Removes a plan. | `result`. |
++| `status` | Reads the status line. | `result` (`Plan '<name>': D/N tasks complete`), `status`. |
++
++| Aspect | Behaviour |
++|--------|-----------|
++| Status line | `[C/N] <icon> Task C: <description> (<status>)` for the task being worked on — the first one in progress, else the first one still open —, `[N/N] Plan finished: D done, F failed` when every task is over and some failed, `[N/N] All tasks complete!` when all are done. Icons: ⏳ open, 🔄 in progress, ✅ done, ❌ failed. |
++| Errors | Returned as text, never as a failed generation: `[error: no plans found for this session]`, `[error: no plan named '<name>' found]`, `[error: plan '<name>' already exists]`, `[error: no task with id N in plan '<name>']`, `[error: unknown action '<action>'. Valid actions: …]`. An `update` without a valid `status` quotes the call back, says that `tasks` is ignored by `update` when it was sent, and gives a call that works. |
++| Storage | In memory, per conversation: the line mode and the window hold one; over ACP each session has its own, and one thread never sees another's plans. Plans are not written to the session file and are lost when mm stops. A conversation that changes starts with no plan: `/new` and `/resume` in every front end, `session/load` and `session/resume` over ACP. |
++| Permission | `create`, `update`, `next` and `delete` are requested when `yolo: false`; a refusal returns `[plan: command rejected by the user]`. `get`, `list` and `status` never ask. |
++| Display | Line mode: `📋 plan: <action> <name>` before the call, then `📋 <status line>` after every action that has one (`create`, `update`, `next`, `status`). Window: the call as a tool call, and `📋 <status line>` as a notice. Editor (ACP): the call as a tool call, the status inside its result. The progress is shown by mm, whether or not the model repeats it. |
++| ACP and window | Tool call of kind `edit` for `create`, `update`, `next`, `delete`, `read` for the others; a result starting with `[error:` is a failed call. |
++
  ## `lsp`
  Declared when `lsp.servers` lists at least one server whose command is on the `PATH` at start-up. A listed server that is not found is skipped with a `[warning: lsp server <name> (<command>) not found on the PATH — skipped]`; when none is found the tool is not declared and the banner says so.

modified docs/fr/README.md +2 -1 Viewed

@@ -26,6 +26,7 @@ Cette documentation est organisée selon les quatre quadrants [Diátaxis](https:
26	26	- [Faire tourner mini-me dans un Docker Sandbox](how-to/run-in-a-docker-sandbox.md) — le kit `mini-me` de [mini-me-kit](https://rickub.com/bots-garden/mini-me-kit), les mixins `go-toolchain` et `cli-tools` sous `sbx/`, mode fenêtre et éditeur
27	27	- [Ajouter un skill](how-to/add-a-skill.md)
28	28	- [Déclarer un outil personnalisé](how-to/declare-a-custom-tool.md)
	29	+- [Gérer des plans et des tâches](how-to/manage-plans-and-tasks.md) — l'outil `plan` : donner une liste, suivre la ligne de progression `📋`, des plans par conversation
29	30	- [Sauvegarder une réponse dans un fichier](how-to/save-an-answer.md)
30	31	- [Chercher dans un dossier de documentation](how-to/search-the-docs.md) — `/docs <mots>`, l'outil `search_docs`, `docSearch` (BM25, sans modèle)
31	32	- [Chercher dans les sessions précédentes](how-to/search-earlier-sessions.md) — `/memory <mots>`, l'outil `search_memory`, `sessions.searchMemory` / `sessions.autoSearchMemory`
@@ -54,7 +55,7 @@ Cette documentation est organisée selon les quatre quadrants [Diátaxis](https:
54	55	- [Ligne de commande](reference/cli.md) — options, variables d'environnement, bannière de démarrage
55	56	- [Commandes slash](reference/slash-commands.md) — `/quit`, `/abort`, `/compact`, `/new`, `/sessions`, `/resume`, `/help`, raccourcis clavier, touches de la fenêtre
56	57	- [Fichier de configuration](reference/configuration.md) — chaque clé YAML, sa valeur par défaut et sa règle de validation
57		-- [Outils intégrés](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, outils déclarés
	58	+- [Outils intégrés](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `plan`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, `search_docs`, `search_memory`, outils déclarés
58	59	- [Format d'un fichier de skill](reference/skill-format.md)
59	60	- [Les hooks](reference/hooks.md) — les six événements, le JSON qu'un hook lit et écrit, `matcher`, `timeout`, `onError`, erreurs de validation
60	61	- [Surface Agent Client Protocol](reference/acp.md)

  - [Faire tourner mini-me dans un Docker Sandbox](how-to/run-in-a-docker-sandbox.md) — le kit `mini-me` de [mini-me-kit](https://rickub.com/bots-garden/mini-me-kit), les mixins `go-toolchain` et `cli-tools` sous `sbx/`, mode fenêtre et éditeur
  - [Ajouter un skill](how-to/add-a-skill.md)
  - [Déclarer un outil personnalisé](how-to/declare-a-custom-tool.md)
++- [Gérer des plans et des tâches](how-to/manage-plans-and-tasks.md) — l'outil `plan` : donner une liste, suivre la ligne de progression `📋`, des plans par conversation
  - [Sauvegarder une réponse dans un fichier](how-to/save-an-answer.md)
  - [Chercher dans un dossier de documentation](how-to/search-the-docs.md) — `/docs <mots>`, l'outil `search_docs`, `docSearch` (BM25, sans modèle)
  - [Chercher dans les sessions précédentes](how-to/search-earlier-sessions.md) — `/memory <mots>`, l'outil `search_memory`, `sessions.searchMemory` / `sessions.autoSearchMemory`
  - [Ligne de commande](reference/cli.md) — options, variables d'environnement, bannière de démarrage
  - [Commandes slash](reference/slash-commands.md) — `/quit`, `/abort`, `/compact`, `/new`, `/sessions`, `/resume`, `/help`, raccourcis clavier, touches de la fenêtre
  - [Fichier de configuration](reference/configuration.md) — chaque clé YAML, sa valeur par défaut et sa règle de validation
--- [Outils intégrés](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, outils déclarés
++- [Outils intégrés](reference/tools.md) — `bash`, `read_skill`, `read_file`, `write_file`, `edit_file`, `plan`, `lsp`, `fetch_content`, `ask_agent`, `ask_user`, `save_answer`, `unload_model`, `search_docs`, `search_memory`, outils déclarés
  - [Format d'un fichier de skill](reference/skill-format.md)
  - [Les hooks](reference/hooks.md) — les six événements, le JSON qu'un hook lit et écrit, `matcher`, `timeout`, `onError`, erreurs de validation
  - [Surface Agent Client Protocol](reference/acp.md)

modified docs/fr/how-to/manage-plans-and-tasks.md +42 -427 Viewed

@@ -1,450 +1,65 @@
1		----
2		-title: Gérer les plans et les tâches
3		----
	1	+# Gérer des plans et des tâches
4	2
5		-# Gérer les plans et les tâches
	3	+Ce guide montre comment faire travailler l'agent sur une liste de tâches avec son outil `plan`, et comment suivre sa progression. L'outil est activé par défaut (`planTool: true`).
6	4
7		-Mini-me inclut un outil intégré plan qui permet à l'agent de créer, suivre et gérer des listes de tâches multi-étapes. Lorsque vous donnez à l'agent une liste de tâches à effectuer, il crée automatiquement un plan et suit la progression, en vous montrant l'état actuel après chaque réponse.
	5	+## Donner une liste à l'agent
8	6
9		-## Exemple rapide
	7	+Demandez plusieurs choses à la fois, sous forme de liste :
10	8
	9	+```text
	10	+Do these 2 things: 1. create a file a.txt containing A, 2. create a file b.txt containing B.
11	11	```
12		-Vous : Fais ces 3 choses : 1. Créer un fichier README, 2. Ajouter une licence, 3. Valider les modifications
13	12
14		-Agent : Je vais m'en occuper en 3 étapes. En commençant par : Créer un fichier README
15		-📋 plan: create user-request-1
	13	+Le modèle crée un plan à partir de la liste, le suit, et met à jour le statut de chaque tâche au fur et à mesure. Les petits modèles le font plus sûrement quand on le leur dit : « Use the plan tool. Do these… ».
16	14
17		-Agent : Fichier README créé.
18		-📋 [2/3] 🔄 Tâche 2 : Ajouter une licence (in_progress)
	15	+## Suivre la progression
19	16
20		-Agent : Licence ajoutée.
21		-📋 [3/3] 🔄 Tâche 3 : Valider les modifications (in_progress)
	17	+mm affiche où en est le plan après chaque action qui change ou lit son état (`create`, `update`, `next`, `status`) — que le modèle en parle ou non dans sa réponse.
22	18
23		-Agent : Toutes les modifications ont été validées.
24		-📋 [3/3] ✅ Toutes les tâches sont terminées !
25		-```
26		-
27		-La ligne `📋` vous montre exactement où se trouve l'agent dans votre plan.
28		-
29		----
30		-
31		-## Fonctionnement
32		-
33		-### Cycle de vie d'un plan
34		-
35		-1. Vous formulez votre demande sous forme de liste de tâches
36		-2. L'agent détecte le plan et le crée en utilisant l'outil `plan`
37		-3. L'agent travaille sur les tâches une par une, en mettant à jour l'état
38		-4. L'agent signale la progression après chaque action
39		-5. Le plan est terminé lorsque toutes les tâches sont terminées
40		-
41		-### Icônes d'état
42		-
43		-\| Icône \| Signification \|
44		-\|-------\|--------------\|
45		-\| ⏳ \| Tâche non démarrée (ouverte) \|
46		-\| 🔄 \| Tâche en cours \|
47		-\| ✅ \| Tâche terminée avec succès \|
48		-\| ❌ \| Tâche en échec \|
49		-
50		-### Format de la ligne d'état
51		-
52		-```
53		-📋 [courante/total] icône Tâche N : description (état)
54		-```
55		-
56		-Exemples :
57		-- `📋 [1/5] ⏳ Tâche 1 : Analyser le code (open)` - Première tâche, non démarrée
58		-- `📋 [2/5] 🔄 Tâche 2 : Corriger les bugs (in_progress)` - Deuxième tâche, en cours
59		-- `📋 [5/5] ✅ Toutes les tâches sont terminées !` - Tout est terminé !
60		-- `📋 [3/5] ❌ Tâche 3 : Exécuter les tests (failed: timeout de connexion)` - Tâche en échec avec erreur
61		-
62		----
63		-
64		-## Scénarios utilisateur
65		-
66		-### Scénario 1 : Liste numérotée
67		-
68		-Votre entrée :
69		-```
70		-Fais ces 5 choses :
71		-1. Vérifier la version actuelle
72		-2. Mettre à jour les dépendances
73		-3. Exécuter les tests
74		-4. Corriger les échecs
75		-5. Déployer en staging
76		-```
77		-
78		-Ce qui se passe :
79		-- L'agent crée un plan avec 5 tâches
80		-- Travaille sur chaque tâche séquentiellement
81		-- Affiche la progression : `📋 [1/5]`, `📋 [2/5]`, etc.
82		-- Termine avec : `📋 [5/5] ✅ Toutes les tâches sont terminées !`
83		-
84		----
85		-
86		-### Scénario 2 : Tâches séparées par des virgules
87		-
88		-Votre entrée :
89		-```
90		-Configurer le projet : créer le répertoire, initialiser git, et ajouter un README
91		-```
92		-
93		-Ce qui se passe :
94		-- L'agent détecte qu'il s'agit d'une demande multi-étapes
95		-- Crée un plan avec 3 tâches
96		-- Exécute dans l'ordre
97		-
98		----
99		-
100		-### Scénario 3 : Plusieurs plans
101		-
102		-Vous pouvez avoir plusieurs plans actifs dans une seule session.
103		-
104		-Votre entrée :
105		-```
106		-D'abord, corrige le module d'authentification : 1. Trouver le bug, 2. Écrire la correction, 3. Le tester
107		-```
108		-
109		-L'agent travaille sur le plan d'authentification, affiche : `📋 [2/3] 🔄 Tâche 2 : Écrire la correction (in_progress)`
110		-
111		-Ensuite vous dites :
112		-```
113		-Aussi, mettre à jour la documentation : 1. Mettre à jour le README, 2. Mettre à jour le CHANGELOG
114		-```
115		-
116		-L'agent crée un deuxième plan et gère les deux indépendamment.
117		-
118		-Pour voir tous les plans actifs :
119		-```
120		-Quels plans ai-je ?
121		-```
122		-
123		----
124		-
125		-## Interagir avec les plans
126		-
127		-### Vérifier l'état actuel du plan
128		-
129		-Demandez : "Quel est l'état de mon plan ?"
130		-
131		-L'agent répond :
132		-```
133		-Plan actuel : [2/3] 🔄 Tâche 2 : Écrire la correction (in_progress). Suivant : Le tester
134		-```
135		-
136		-### Sauter une tâche
137		-
138		-Demandez : "Saute l'étape de test, déploie directement"
139		-
140		-Agent :
141		-- Marque la tâche actuelle comme terminée
142		-- Saute à la tâche suivante
143		-- Confirme : "Je saute l'étape de test. Déploiement en staging..."
144		-
145		-### Abandonner un plan
146		-
147		-Demandez : "Peu importe, je n'en ai plus besoin"
148		-
149		-Agent :
150		-- Demande confirmation
151		-- Si vous confirmez, supprime le plan
152		-- Toute la progression est perdue
153		-
154		----
155		-
156		-## Commandes de gestion des plans
157		-
158		-Bien que l'outil plan soit principalement piloté par le modèle (l'agent l'utilise automatiquement), vous pouvez également interagir directement avec les plans :
159		-
160		-### Lister tous les plans
161		-
162		-Commande : `/plans` (pas encore implémenté en v1 - arrive en v2)
163		-
164		-Ce que cela affiche :
165		-```
166		-Plans actifs :
167		-1. configuration-projet (2/5 tâches terminées)
168		-2. corriger-bugs (0/3 tâches terminées)
169		-```
170		-
171		-### Voir les détails d'un plan
172		-
173		-Demandez à l'agent : "Montre-moi les détails du plan X"
174		-
175		-L'agent répond : Liste complète des tâches avec l'état pour chaque tâche
176		-
177		-### Supprimer un plan
178		-
179		-Demandez à l'agent : "Supprime le plan X"
180		-
181		-Agent : Confirme et supprime le plan
182		-
183		----
184		-
185		-## Cycle de vie de la session
186		-
187		-### Plans et commande /new
188		-
189		-Lorsque vous démarrez une nouvelle session avec `/new` :
190		-- Tous les plans actifs sont supprimés
191		-- C'est intentionnel : une nouvelle session signifie un nouveau départ
192		-- Si vous voulez continuer un ancien plan, ne pas utiliser `/new`
193		-
194		-### Plans et redémarrage de l'agent
195		-
196		-En v1 (version actuelle) :
197		-- Les plans sont stockés uniquement en mémoire
198		-- Les plans sont perdus lorsque l'agent redémarre
199		-- Cela garde la v1 simple et rapide
200		-
201		-En v2 (future) :
202		-- Les plans seront persistés dans les fichiers de session
203		-- Les plans survivront aux redémarrages de l'agent
204		-
205		----
206		-
207		-## L'outil plan (pour les utilisateurs avancés)
208		-
209		-L'outil `plan` est disponible pour que le modèle puisse l'appeler. Il prend en charge ces actions :
210		-
211		-\| Action \| Objectif \| Paramètres requis \|
212		-\|--------\|----------\|-------------------\|
213		-\| `create` \| Créer un nouveau plan \| `tasks` (tableau de `{description}`) \|
214		-\| `get` \| Obtenir les détails d'un plan \| `name` \|
215		-\| `update` \| Mettre à jour l'état d'une tâche \| `name`, `taskId`, `status` \|
216		-\| `next` \| Passer à la tâche suivante \| `name` \|
217		-\| `list` \| Lister tous les plans \| aucun \|
218		-\| `delete` \| Supprimer un plan \| `name` \|
219		-\| `status` \| Obtenir la ligne d'état \| `name` \|
220		-
221		-### Valeurs d'état
222		-
223		-- `open` - Tâche non démarrée
224		-- `in_progress` - Tâche actuellement en cours
225		-- `done` - Tâche terminée avec succès
226		-- `failed` - Tâche en échec
227		-
228		----
229		-
230		-## La compétence plan-manager
231		-
232		-L'agent utilise la compétence `plan-manager` pour savoir quand et comment utiliser l'outil plan. Cette compétence inclut :
233		-
234		-- Modèles de détection : Comment reconnaître quand un utilisateur donne une demande multi-étapes
235		-- Directives d'utilisation : Quand créer des plans, mettre à jour des tâches, passer entre les tâches
236		-- Format d'affichage : Comment montrer l'état du plan à l'utilisateur
237		-- Gestion des erreurs : Que faire quand les choses tournent mal
238		-
239		-Vous n'avez pas besoin de charger manuellement cette compétence - l'agent le fait automatiquement lorsque cela est approprié.
240		-
241		----
242		-
243		-## Exemples
244		-
245		-### Exemple 1 : Workflow de correction de bug
246		-
247		-```
248		-Vous : Corrige ces problèmes dans le module d'authentification :
249		- 1. Vérifier les exceptions de pointeur nul
250		- 2. Ajouter la validation des entrées
251		- 3. Écrire des tests unitaires
252		- 4. Mettre à jour la documentation
253		-
254		-Agent : Je vais m'en occuper en 4 étapes. En commençant par : Vérifier les exceptions de pointeur nul
255		-📋 plan: create user-request-1
256		-
257		-... (l'agent vérifie le code) ...
258		-
259		-Agent : 3 problèmes de pointeur nul potentiels trouvés.
260		-📋 [2/4] 🔄 Tâche 2 : Ajouter la validation des entrées (in_progress)
261		-
262		-... (l'agent ajoute la validation) ...
263		-
264		-Agent : Validation des entrées ajoutée à auth.go.
265		-📋 [3/4] 🔄 Tâche 3 : Écrire des tests unitaires (in_progress)
266		-
267		-... (l'agent écrit les tests) ...
268		-
269		-Agent : Tests écrits et réussis.
270		-📋 [4/4] 🔄 Tâche 4 : Mettre à jour la documentation (in_progress)
271		-
272		-... (l'agent met à jour les docs) ...
273		-
274		-Agent : Documentation mise à jour avec les nouvelles exigences de validation.
275		-📋 [4/4] ✅ Toutes les tâches sont terminées !
276		-```
277		-
278		----
279		-
280		-### Exemple 2 : Configuration d'un projet
281		-
282		-```
283		-Vous : Configure un nouveau projet Go avec :
284		- 1. Initialisation du module
285		- 2. Structure des répertoires
286		- 3. Fichier main
287		- 4. Fichier de test
288		-
289		-Agent : Je vais configurer cela en 4 étapes. En commençant par : Initialisation du module
290		-📋 plan: create user-request-1
291		-
292		-Agent : Module initialisé avec go mod
293		-📋 [2/4] 🔄 Tâche 2 : Structure des répertoires (in_progress)
294		-
295		-Agent : Répertoires créés (cmd, internal, pkg, tests)
296		-📋 [3/4] 🔄 Tâche 3 : Fichier main (in_progress)
297		-
298		-Agent : Fichier main créé avec la structure de base
299		-📋 [4/4] 🔄 Tâche 4 : Fichier de test (in_progress)
300		-
301		-Agent : Fichier de test créé avec des tests exemples
302		-📋 [4/4] ✅ Toutes les tâches sont terminées !
303		-```
304		-
305		----
306		-
307		-### Exemple 3 : Revue de code
308		-
309		-```
310		-Vous : Revoir cette PR et :
311		- 1. Vérifier les problèmes de sécurité
312		- 2. Vérifier que les tests passent
313		- 3. Revoir la documentation
314		- 4. Laisser des commentaires
315		-
316		-Agent : Je vais revoir cette PR en 4 étapes. En commençant par : Vérifier les problèmes de sécurité
317		-📋 plan: create user-request-1
318		-
319		-Agent : Aucun problème de sécurité trouvé.
320		-📋 [2/4] 🔄 Tâche 2 : Vérifier que les tests passent (in_progress)
321		-
322		-Agent : Tous les tests passent.
323		-📋 [3/4] 🔄 Tâche 3 : Revoir la documentation (in_progress)
324		-
325		-Agent : La documentation a l'air bonne, il faut juste corriger une faute de frappe.
326		-📋 [4/4] 🔄 Tâche 4 : Laisser des commentaires (in_progress)
327		-
328		-Agent : Commentaires laissés sur la PR.
329		-📋 [4/4] ✅ Toutes les tâches sont terminées !
330		-```
331		-
332		----
333		-
334		-### Exemple 4 : Avec un échec de tâche
335		-
336		-```
337		-Vous : Fais ceci :
338		- 1. Se connecter à la base de données
339		- 2. Exécuter la migration
340		- 3. Importer les données
341		-
342		-Agent : Je vais m'en occuper en 3 étapes. En commençant par : Se connecter à la base de données
343		-📋 plan: create user-request-1
344		-
345		-Agent : Connexion à la base de données établie.
346		-📋 [2/3] 🔄 Tâche 2 : Exécuter la migration (in_progress)
347		-
348		-Agent : Migration terminée.
349		-📋 [3/3] 🔄 Tâche 3 : Importer les données (in_progress)
350		-
351		-Agent : Erreur : timeout de connexion à la source de données
352		-📋 plan: update user-request-1
353		-
354		-Agent : ⚠️ La tâche 3 a échoué : timeout de connexion à la source de données
355		-📋 [3/3] ❌ Tâche 3 : Importer les données (failed: connection timeout)
356		-
357		-Agent : 2 des 3 tâches sont terminées. L'import des données a échoué en raison d'un timeout de connexion. Voulez-vous que je réessaye, ou que je saute cette tâche ?
358		-```
359		-
360		----
361		-
362		-## Configuration
363		-
364		-L'outil plan peut être activé ou désactivé via le fichier de configuration YAML :
365		-
366		-```yaml
367		-# Dans votre fichier agent.yaml
368		-planTool: true # Activé par défaut
369		-```
370		-
371		-Pour désactiver :
372		-```yaml
373		-planTool: false
374		-```
375		-
376		-Lorsque désactivé, l'agent n'aura pas accès à l'outil `plan` et ne créera pas automatiquement de plans.
377		-
378		----
379		-
380		-## Résolution des problèmes
381		-
382		-### L'agent ne crée pas de plans
383		-
384		-Si l'agent ne crée pas de plans pour vos demandes multi-étapes :
385		-
386		-1. Rendez votre demande plus claire - Utilisez des listes numérotées
387		- - ❌ "Corrige quelques choses"
388		- - ✅ "Fais ceci : 1. Corrige A, 2. Corrige B, 3. Corrige C"
389		-
390		-2. Soyez plus explicite - Utilisez des mots comme "plan", "étapes", "puis"
391		- - ✅ "Mon plan est de d'abord faire X, puis faire Y"
392		-
393		-3. Vérifiez que la compétence est chargée - L'agent devrait automatiquement charger la compétence `plan-manager` lorsqu'il détecte une demande de type plan
394		-
395		-### L'agent appelle `plan update` en boucle
396		-
397		-`update` change l'état d'une tâche ; il ne peut pas modifier les tâches d'un plan. Un modèle qui envoie un `update` sans `status` valide — souvent avec une liste `tasks`, pour essayer de modifier le plan — reçoit une réponse qui cite son appel, nomme le champ manquant, précise que `tasks` n'est lu que par `create`, et donne un appel valide :
398		-
399		-```
400		-[error: plan update changes the status of ONE task, and "status" is missing. You sent: {…}. "status" must be one of: open, in_progress, done, failed. "tasks" is ignored by update (only create reads it): to change a plan's tasks, delete the plan and create it again. A valid call: {"action":"update","name":"user-request-1","taskId":1,"status":"done"}]
401		-```
402		-
403		-Si un petit modèle répète quand même l'appel, mettez `planTool: false` pour un agent dont le métier n'a pas besoin de plans (écrire un document, répondre à des questions) : le modèle travaille alors sans.
404		-
405		-### Les plans disparaissent
406		-
407		-En v1, les plans sont stockés en mémoire et seront perdus lorsque :
408		-- Vous démarrez une nouvelle session avec `/new`
409		-- L'agent redémarre
410		-- Vous fermez et rouvrez le terminal
411		-
412		-C'est un comportement attendu en v1. La persistance des sessions arrive en v2.
413		-
414		----
	19	+- Mode ligne (`mm agent.yaml`, `mm -p`) : une ligne avant chaque appel, une ligne après. Mesuré sur Qwen3.5 4B avec la demande ci-dessus :
415	20
416		-## Bonnes pratiques
	21	+ ```text
	22	+ 📋 plan: create
	23	+ 📋 [1/2] ⏳ Task 1: Create a file a.txt containing A (open)
	24	+ 💾 write_file: a.txt (1 bytes)
	25	+ 💾 write_file: b.txt (1 bytes)
	26	+ 📋 plan: update user-request-1
	27	+ 📋 [2/2] ⏳ Task 2: Create a file b.txt containing B (open)
	28	+ 📋 plan: update user-request-1
	29	+ 📋 [2/2] All tasks complete!
	30	+ ```
417	31
418		-1. Soyez spécifique - Plus il y a de détails dans vos descriptions de tâches, mieux l'agent peut les exécuter
	32	+- Fenêtre (`mm -tui`) : chaque appel est un appel d'outil dans la conversation, et la ligne `📋 [C/N] …` le suit en notice.
	33	+- Éditeur (`mm -acp`) : chaque appel est un appel d'outil ; la ligne de statut est dans son résultat.
419	34
420		-2. Gardez les plans concentrés - Un plan par demande logique. Ne mélangez pas des tâches sans rapport
	35	+La ligne de statut montre la tâche en cours de travail — la première en cours, sinon la première encore ouverte : `[C/N] <icône> Task C: <description> (<statut>)`. Icônes : ⏳ ouverte, 🔄 en cours, ✅ faite, ❌ en échec. Quand toutes les tâches sont terminées, elle indique `[N/N] All tasks complete!`, ou `[N/N] Plan finished: D done, F failed` si certaines ont échoué.
421	36
422		-3. Vérifiez la progression - La ligne d'état vous montre exactement où se trouve l'agent
	37	+## Orienter le plan
423	38
424		-4. Demandez des clarifications - Si l'agent semble bloqué ou confus, demandez "Quel est l'état ?"
	39	+C'est le modèle qui décide quand appeler l'outil ; vous l'orientez avec vos propres mots :
425	40
426		-5. Commencez petit - Pour un travail multi-étapes complexe, commencez par quelques tâches et ajoutez-en davantage si nécessaire
	41	+- Où en est-on ? « Où en est le plan ? » — le modèle appelle `status` et mm affiche la ligne.
	42	+- Sauter une tâche. « Saute les tests et continue. » — le modèle peut la marquer et passer à la suivante avec `next`.
	43	+- Une tâche a échoué. Le modèle peut la marquer `failed`, avec une note ; la ligne de statut passe alors à la tâche ouverte suivante.
	44	+- Abandonner le plan. « Oublie ce plan. » — le modèle peut appeler `delete`.
427	45
428		----
	46	+Il n'y a pas de commande slash pour les plans : c'est l'outil du modèle.
429	47
430		-## Comparaison avec d'autres outils
	48	+## Variantes
431	49
432		-\| Fonctionnalité \| outil plan \| compétences \| commandes slash \|
433		-\|---------------\|------------\|-------------\|-----------------\|
434		-\| Objectif \| Suivre le travail multi-étapes \| Charger des procédures \| Commandes utilisateur \|
435		-\| Déclenchement \| Automatique (modèle détecte) \| Explicite (`/skill`) \| Explicite (`/command`) \|
436		-\| Persistance (v1) \| En mémoire, session uniquement \| Basé sur des fichiers \| N/A \|
437		-\| Interaction utilisateur \| Indirecte (via le modèle) \| Indirecte (modèle charge) \| Directe \|
	50	+- Plusieurs conversations dans un éditeur. Chaque fil a ses propres plans : un plan créé dans un fil de Zed n'est pas visible dans un autre.
	51	+- Repartir de zéro. `/new` et `/resume` oublient les plans dans toutes les façades ; un rechargement depuis l'éditeur (`session/load`, `session/resume`) démarre le fil sans plan. Les plans ne sont gardés qu'en mémoire, jamais dans le fichier de session, et sont perdus à l'arrêt de mm.
	52	+- Le skill `plan-manager`. `skills/plan-manager` dit au modèle quand faire un plan et comment le parcourir. Mettez-le dans le `skillsDir` de l'agent pour un modèle qui n'utilise pas l'outil de lui-même. Il demande aussi au modèle de reprendre la ligne de statut dans sa réponse ; mm l'affiche de toute façon.
	53	+- Un agent qui n'a pas besoin de plans. Mettez `planTool: false` : l'outil n'est pas déclaré. Faites-le pour un agent dont le travail est une réponse unique (écrire un document, répondre à des questions), où un petit modèle pourrait sinon boucler sur des appels au plan.
438	54
439		-L'outil `plan` complète les autres fonctionnalités de mini-me :
440		-- Utilisez les compétences pour des procédures complexes et réutilisables
441		-- Utilisez l'outil plan pour des demandes multi-étapes ad-hoc
442		-- Utilisez les commandes slash pour des actions utilisateur directes
	55	+## Dépannage
443	56
444		----
	57	+- L'agent ne crée pas de plan. Donnez une liste numérotée, et dites « use the plan tool ». Avec un petit modèle, ajoutez le skill `plan-manager`.
	58	+- L'agent appelle `plan update` en boucle. `update` change le statut d'UNE tâche ; il ne peut pas changer les tâches d'un plan. Un appel sans `status` valide reçoit une réponse qui reprend l'appel, nomme le champ manquant et donne un appel valide. Si un petit modèle le répète quand même, mettez `planTool: false` pour cet agent.
445	59
446	60	## Voir aussi
447	61
448		-- [Référence des commandes slash](../reference/slash-commands.md) - Pour les commandes utilisateur
449		-- [Référence des outils](../reference/tools.md) - Pour tous les outils intégrés
450		-- [Aperçu des compétences](../explanation/skills.md) - Pour le système de compétences
	62	+- [Référence des outils — `plan`](../reference/tools.md#plan) : chaque action, champ, résultat et erreur
	63	+- [Référence de la configuration](../reference/configuration.md) : `planTool`
	64	+- [Référence des commandes](../reference/slash-commands.md) : `/new`, `/resume`
	65	+- `demos/plans/` — une démo mesurée : un programme Go construit en quatre étapes planifiées, et un plan qui se termine sur une tâche en échec

-----
++# Gérer des plans et des tâches
--title: Gérer les plans et les tâches
-----
--# Gérer les plans et les tâches
++Ce guide montre comment faire travailler l'agent sur une liste de tâches avec son outil `plan`, et comment suivre sa progression. L'outil est activé par défaut (`planTool: true`).
--Mini-me inclut un outil intégré **plan** qui permet à l'agent de créer, suivre et gérer des listes de tâches multi-étapes. Lorsque vous donnez à l'agent une liste de tâches à effectuer, il crée automatiquement un plan et suit la progression, en vous montrant l'état actuel après chaque réponse.
++## Donner une liste à l'agent
--## Exemple rapide
++Demandez plusieurs choses à la fois, sous forme de liste :
++```text
++Do these 2 things: 1. create a file a.txt containing A, 2. create a file b.txt containing B.
  ```
--Vous : Fais ces 3 choses : 1. Créer un fichier README, 2. Ajouter une licence, 3. Valider les modifications
--Agent : Je vais m'en occuper en 3 étapes. En commençant par : Créer un fichier README
++Le modèle crée un plan à partir de la liste, le suit, et met à jour le statut de chaque tâche au fur et à mesure. Les petits modèles le font plus sûrement quand on le leur dit : « Use the plan tool. Do these… ».
--📋 plan: create user-request-1
--Agent : Fichier README créé.
++## Suivre la progression
--📋 [2/3] 🔄 Tâche 2 : Ajouter une licence (in_progress)
--Agent : Licence ajoutée.
++mm affiche où en est le plan après chaque action qui change ou lit son état (`create`, `update`, `next`, `status`) — que le modèle en parle ou non dans sa réponse.
--📋 [3/3] 🔄 Tâche 3 : Valider les modifications (in_progress)
--Agent : Toutes les modifications ont été validées.
++- **Mode ligne** (`mm agent.yaml`, `mm -p`) : une ligne avant chaque appel, une ligne après. Mesuré sur Qwen3.5 4B avec la demande ci-dessus :
--📋 [3/3] ✅ Toutes les tâches sont terminées !
--```
--
--La ligne `📋` vous montre exactement où se trouve l'agent dans votre plan.
--
-----
--
--## Fonctionnement
--
--### Cycle de vie d'un plan
--
--1. **Vous formulez votre demande** sous forme de liste de tâches
--2. **L'agent détecte le plan** et le crée en utilisant l'outil `plan`
--3. **L'agent travaille sur les tâches** une par une, en mettant à jour l'état
--4. **L'agent signale la progression** après chaque action
--5. **Le plan est terminé** lorsque toutes les tâches sont terminées
--
--### Icônes d'état
--
--| Icône | Signification |
--|-------|--------------|
--| ⏳ | Tâche non démarrée (ouverte) |
--| 🔄 | Tâche en cours |
--| ✅ | Tâche terminée avec succès |
--| ❌ | Tâche en échec |
--
--### Format de la ligne d'état
--
--```
--📋 [courante/total] icône Tâche N : description (état)
--```
--
--Exemples :
--- `📋 [1/5] ⏳ Tâche 1 : Analyser le code (open)` - Première tâche, non démarrée
--- `📋 [2/5] 🔄 Tâche 2 : Corriger les bugs (in_progress)` - Deuxième tâche, en cours
--- `📋 [5/5] ✅ Toutes les tâches sont terminées !` - Tout est terminé !
--- `📋 [3/5] ❌ Tâche 3 : Exécuter les tests (failed: timeout de connexion)` - Tâche en échec avec erreur
--
-----
--
--## Scénarios utilisateur
--
--### Scénario 1 : Liste numérotée
--
--**Votre entrée :**
--```
--Fais ces 5 choses :
--1. Vérifier la version actuelle
--2. Mettre à jour les dépendances
--3. Exécuter les tests
--4. Corriger les échecs
--5. Déployer en staging
--```
--
--**Ce qui se passe :**
--- L'agent crée un plan avec 5 tâches
--- Travaille sur chaque tâche séquentiellement
--- Affiche la progression : `📋 [1/5]`, `📋 [2/5]`, etc.
--- Termine avec : `📋 [5/5] ✅ Toutes les tâches sont terminées !`
--
-----
--
--### Scénario 2 : Tâches séparées par des virgules
--
--**Votre entrée :**
--```
--Configurer le projet : créer le répertoire, initialiser git, et ajouter un README
--```
--
--**Ce qui se passe :**
--- L'agent détecte qu'il s'agit d'une demande multi-étapes
--- Crée un plan avec 3 tâches
--- Exécute dans l'ordre
--
-----
--
--### Scénario 3 : Plusieurs plans
--
--Vous pouvez avoir plusieurs plans actifs dans une seule session.
--
--**Votre entrée :**
--```
--D'abord, corrige le module d'authentification : 1. Trouver le bug, 2. Écrire la correction, 3. Le tester
--```
--
--L'agent travaille sur le plan d'authentification, affiche : `📋 [2/3] 🔄 Tâche 2 : Écrire la correction (in_progress)`
--
--**Ensuite vous dites :**
--```
--Aussi, mettre à jour la documentation : 1. Mettre à jour le README, 2. Mettre à jour le CHANGELOG
--```
--
--L'agent crée un deuxième plan et gère les deux indépendamment.
--
--**Pour voir tous les plans actifs :**
--```
--Quels plans ai-je ?
--```
--
-----
--
--## Interagir avec les plans
--
--### Vérifier l'état actuel du plan
--
--**Demandez :** "Quel est l'état de mon plan ?"
--
--**L'agent répond :**
--```
--Plan actuel : [2/3] 🔄 Tâche 2 : Écrire la correction (in_progress). Suivant : Le tester
--```
--
--### Sauter une tâche
--
--**Demandez :** "Saute l'étape de test, déploie directement"
--
--**Agent :**
--- Marque la tâche actuelle comme terminée
--- Saute à la tâche suivante
--- Confirme : "Je saute l'étape de test. Déploiement en staging..."
--
--### Abandonner un plan
--
--**Demandez :** "Peu importe, je n'en ai plus besoin"
--
--**Agent :**
--- Demande confirmation
--- Si vous confirmez, supprime le plan
--- Toute la progression est perdue
--
-----
--
--## Commandes de gestion des plans
--
--Bien que l'outil plan soit principalement piloté par le modèle (l'agent l'utilise automatiquement), vous pouvez également interagir directement avec les plans :
--
--### Lister tous les plans
--
--**Commande :** `/plans` (pas encore implémenté en v1 - arrive en v2)
--
--**Ce que cela affiche :**
--```
--Plans actifs :
--1. configuration-projet (2/5 tâches terminées)
--2. corriger-bugs (0/3 tâches terminées)
--```
--
--### Voir les détails d'un plan
--
--**Demandez à l'agent :** "Montre-moi les détails du plan X"
--
--**L'agent répond :** Liste complète des tâches avec l'état pour chaque tâche
--
--### Supprimer un plan
--
--**Demandez à l'agent :** "Supprime le plan X"
--
--**Agent :** Confirme et supprime le plan
--
-----
--
--## Cycle de vie de la session
--
--### Plans et commande /new
--
--Lorsque vous démarrez une nouvelle session avec `/new` :
--- Tous les plans actifs sont **supprimés**
--- C'est intentionnel : une nouvelle session signifie un nouveau départ
--- Si vous voulez continuer un ancien plan, ne pas utiliser `/new`
--
--### Plans et redémarrage de l'agent
--
--En v1 (version actuelle) :
--- Les plans sont stockés **uniquement en mémoire**
--- Les plans sont **perdus** lorsque l'agent redémarre
--- Cela garde la v1 simple et rapide
--
--En v2 (future) :
--- Les plans seront persistés dans les fichiers de session
--- Les plans survivront aux redémarrages de l'agent
--
-----
--
--## L'outil plan (pour les utilisateurs avancés)
--
--L'outil `plan` est disponible pour que le modèle puisse l'appeler. Il prend en charge ces actions :
--
--| Action | Objectif | Paramètres requis |
--|--------|----------|-------------------|
--| `create` | Créer un nouveau plan | `tasks` (tableau de `{description}`) |
--| `get` | Obtenir les détails d'un plan | `name` |
--| `update` | Mettre à jour l'état d'une tâche | `name`, `taskId`, `status` |
--| `next` | Passer à la tâche suivante | `name` |
--| `list` | Lister tous les plans | aucun |
--| `delete` | Supprimer un plan | `name` |
--| `status` | Obtenir la ligne d'état | `name` |
--
--### Valeurs d'état
--
--- `open` - Tâche non démarrée
--- `in_progress` - Tâche actuellement en cours
--- `done` - Tâche terminée avec succès
--- `failed` - Tâche en échec
--
-----
--
--## La compétence plan-manager
--
--L'agent utilise la compétence `plan-manager` pour savoir **quand et comment** utiliser l'outil plan. Cette compétence inclut :
--
--- **Modèles de détection** : Comment reconnaître quand un utilisateur donne une demande multi-étapes
--- **Directives d'utilisation** : Quand créer des plans, mettre à jour des tâches, passer entre les tâches
--- **Format d'affichage** : Comment montrer l'état du plan à l'utilisateur
--- **Gestion des erreurs** : Que faire quand les choses tournent mal
--
--Vous n'avez pas besoin de charger manuellement cette compétence - l'agent le fait automatiquement lorsque cela est approprié.
--
-----
--
--## Exemples
--
--### Exemple 1 : Workflow de correction de bug
--
--```
--Vous : Corrige ces problèmes dans le module d'authentification :
--     1. Vérifier les exceptions de pointeur nul
--     2. Ajouter la validation des entrées
--     3. Écrire des tests unitaires
--     4. Mettre à jour la documentation
--
--Agent : Je vais m'en occuper en 4 étapes. En commençant par : Vérifier les exceptions de pointeur nul
--📋 plan: create user-request-1
--
--... (l'agent vérifie le code) ...
--
--Agent : 3 problèmes de pointeur nul potentiels trouvés.
--📋 [2/4] 🔄 Tâche 2 : Ajouter la validation des entrées (in_progress)
--
--... (l'agent ajoute la validation) ...
--
--Agent : Validation des entrées ajoutée à auth.go.
--📋 [3/4] 🔄 Tâche 3 : Écrire des tests unitaires (in_progress)
--
--... (l'agent écrit les tests) ...
--
--Agent : Tests écrits et réussis.
--📋 [4/4] 🔄 Tâche 4 : Mettre à jour la documentation (in_progress)
--
--... (l'agent met à jour les docs) ...
--
--Agent : Documentation mise à jour avec les nouvelles exigences de validation.
--📋 [4/4] ✅ Toutes les tâches sont terminées !
--```
--
-----
--
--### Exemple 2 : Configuration d'un projet
--
--```
--Vous : Configure un nouveau projet Go avec :
--     1. Initialisation du module
--     2. Structure des répertoires
--     3. Fichier main
--     4. Fichier de test
--
--Agent : Je vais configurer cela en 4 étapes. En commençant par : Initialisation du module
--📋 plan: create user-request-1
--
--Agent : Module initialisé avec go mod
--📋 [2/4] 🔄 Tâche 2 : Structure des répertoires (in_progress)
--
--Agent : Répertoires créés (cmd, internal, pkg, tests)
--📋 [3/4] 🔄 Tâche 3 : Fichier main (in_progress)
--
--Agent : Fichier main créé avec la structure de base
--📋 [4/4] 🔄 Tâche 4 : Fichier de test (in_progress)
--
--Agent : Fichier de test créé avec des tests exemples
--📋 [4/4] ✅ Toutes les tâches sont terminées !
--```
--
-----
--
--### Exemple 3 : Revue de code
--
--```
--Vous : Revoir cette PR et :
--     1. Vérifier les problèmes de sécurité
--     2. Vérifier que les tests passent
--     3. Revoir la documentation
--     4. Laisser des commentaires
--
--Agent : Je vais revoir cette PR en 4 étapes. En commençant par : Vérifier les problèmes de sécurité
--📋 plan: create user-request-1
--
--Agent : Aucun problème de sécurité trouvé.
--📋 [2/4] 🔄 Tâche 2 : Vérifier que les tests passent (in_progress)
--
--Agent : Tous les tests passent.
--📋 [3/4] 🔄 Tâche 3 : Revoir la documentation (in_progress)
--
--Agent : La documentation a l'air bonne, il faut juste corriger une faute de frappe.
--📋 [4/4] 🔄 Tâche 4 : Laisser des commentaires (in_progress)
--
--Agent : Commentaires laissés sur la PR.
--📋 [4/4] ✅ Toutes les tâches sont terminées !
--```
--
-----
--
--### Exemple 4 : Avec un échec de tâche
--
--```
--Vous : Fais ceci :
--     1. Se connecter à la base de données
--     2. Exécuter la migration
--     3. Importer les données
--
--Agent : Je vais m'en occuper en 3 étapes. En commençant par : Se connecter à la base de données
--📋 plan: create user-request-1
--
--Agent : Connexion à la base de données établie.
--📋 [2/3] 🔄 Tâche 2 : Exécuter la migration (in_progress)
--
--Agent : Migration terminée.
--📋 [3/3] 🔄 Tâche 3 : Importer les données (in_progress)
--
--Agent : Erreur : timeout de connexion à la source de données
--📋 plan: update user-request-1
--
--Agent : ⚠️ La tâche 3 a échoué : timeout de connexion à la source de données
--📋 [3/3] ❌ Tâche 3 : Importer les données (failed: connection timeout)
--
--Agent : 2 des 3 tâches sont terminées. L'import des données a échoué en raison d'un timeout de connexion. Voulez-vous que je réessaye, ou que je saute cette tâche ?
--```
--
-----
--
--## Configuration
--
--L'outil plan peut être activé ou désactivé via le fichier de configuration YAML :
--
--```yaml
--# Dans votre fichier agent.yaml
--planTool: true  # Activé par défaut
--```
--
--Pour désactiver :
--```yaml
--planTool: false
--```
--
--Lorsque désactivé, l'agent n'aura pas accès à l'outil `plan` et ne créera pas automatiquement de plans.
--
-----
--
--## Résolution des problèmes
--
--### L'agent ne crée pas de plans
--
--Si l'agent ne crée pas de plans pour vos demandes multi-étapes :
--
--1. **Rendez votre demande plus claire** - Utilisez des listes numérotées
--   - ❌ "Corrige quelques choses"
--   - ✅ "Fais ceci : 1. Corrige A, 2. Corrige B, 3. Corrige C"
--
--2. **Soyez plus explicite** - Utilisez des mots comme "plan", "étapes", "puis"
--   - ✅ "Mon plan est de d'abord faire X, puis faire Y"
--
--3. **Vérifiez que la compétence est chargée** - L'agent devrait automatiquement charger la compétence `plan-manager` lorsqu'il détecte une demande de type plan
--
--### L'agent appelle `plan update` en boucle
--
--`update` change l'état d'une tâche ; il ne peut pas modifier les tâches d'un plan. Un modèle qui envoie un `update` sans `status` valide — souvent avec une liste `tasks`, pour essayer de modifier le plan — reçoit une réponse qui cite son appel, nomme le champ manquant, précise que `tasks` n'est lu que par `create`, et donne un appel valide :
--
--```
--[error: plan update changes the status of ONE task, and "status" is missing. You sent: {…}. "status" must be one of: open, in_progress, done, failed. "tasks" is ignored by update (only create reads it): to change a plan's tasks, delete the plan and create it again. A valid call: {"action":"update","name":"user-request-1","taskId":1,"status":"done"}]
--```
--
--Si un petit modèle répète quand même l'appel, mettez `planTool: false` pour un agent dont le métier n'a pas besoin de plans (écrire un document, répondre à des questions) : le modèle travaille alors sans.
--
--### Les plans disparaissent
--
--En v1, les plans sont stockés en mémoire et seront perdus lorsque :
--- Vous démarrez une nouvelle session avec `/new`
--- L'agent redémarre
--- Vous fermez et rouvrez le terminal
--
--C'est un comportement attendu en v1. La persistance des sessions arrive en v2.
--
-----
--## Bonnes pratiques
++  ```text
++  📋 plan: create
++  📋 [1/2] ⏳ Task 1: Create a file a.txt containing A (open)
++  💾 write_file: a.txt (1 bytes)
++  💾 write_file: b.txt (1 bytes)
++  📋 plan: update user-request-1
++  📋 [2/2] ⏳ Task 2: Create a file b.txt containing B (open)
++  📋 plan: update user-request-1
++  📋 [2/2] All tasks complete!
++  ```
--1. **Soyez spécifique** - Plus il y a de détails dans vos descriptions de tâches, mieux l'agent peut les exécuter
++- **Fenêtre** (`mm -tui`) : chaque appel est un appel d'outil dans la conversation, et la ligne `📋 [C/N] …` le suit en notice.
++- **Éditeur** (`mm -acp`) : chaque appel est un appel d'outil ; la ligne de statut est dans son résultat.
--2. **Gardez les plans concentrés** - Un plan par demande logique. Ne mélangez pas des tâches sans rapport
++La ligne de statut montre la tâche en cours de travail — la première en cours, sinon la première encore ouverte : `[C/N] <icône> Task C: <description> (<statut>)`. Icônes : ⏳ ouverte, 🔄 en cours, ✅ faite, ❌ en échec. Quand toutes les tâches sont terminées, elle indique `[N/N] All tasks complete!`, ou `[N/N] Plan finished: D done, F failed` si certaines ont échoué.
--3. **Vérifiez la progression** - La ligne d'état vous montre exactement où se trouve l'agent
++## Orienter le plan
--4. **Demandez des clarifications** - Si l'agent semble bloqué ou confus, demandez "Quel est l'état ?"
++C'est le modèle qui décide quand appeler l'outil ; vous l'orientez avec vos propres mots :
--5. **Commencez petit** - Pour un travail multi-étapes complexe, commencez par quelques tâches et ajoutez-en davantage si nécessaire
++- **Où en est-on ?** « Où en est le plan ? » — le modèle appelle `status` et mm affiche la ligne.
++- **Sauter une tâche.** « Saute les tests et continue. » — le modèle peut la marquer et passer à la suivante avec `next`.
++- **Une tâche a échoué.** Le modèle peut la marquer `failed`, avec une note ; la ligne de statut passe alors à la tâche ouverte suivante.
++- **Abandonner le plan.** « Oublie ce plan. » — le modèle peut appeler `delete`.
-----
++Il n'y a pas de commande slash pour les plans : c'est l'outil du modèle.
--## Comparaison avec d'autres outils
++## Variantes
--| Fonctionnalité | outil plan | compétences | commandes slash |
++- **Plusieurs conversations dans un éditeur.** Chaque fil a ses propres plans : un plan créé dans un fil de Zed n'est pas visible dans un autre.
--|---------------|------------|-------------|-----------------|
++- **Repartir de zéro.** `/new` et `/resume` oublient les plans dans toutes les façades ; un rechargement depuis l'éditeur (`session/load`, `session/resume`) démarre le fil sans plan. Les plans ne sont gardés qu'en mémoire, jamais dans le fichier de session, et sont perdus à l'arrêt de mm.
--| Objectif | Suivre le travail multi-étapes | Charger des procédures | Commandes utilisateur |
++- **Le skill `plan-manager`.** `skills/plan-manager` dit au modèle quand faire un plan et comment le parcourir. Mettez-le dans le `skillsDir` de l'agent pour un modèle qui n'utilise pas l'outil de lui-même. Il demande aussi au modèle de reprendre la ligne de statut dans sa réponse ; mm l'affiche de toute façon.
--| Déclenchement | Automatique (modèle détecte) | Explicite (`/skill`) | Explicite (`/command`) |
++- **Un agent qui n'a pas besoin de plans.** Mettez `planTool: false` : l'outil n'est pas déclaré. Faites-le pour un agent dont le travail est une réponse unique (écrire un document, répondre à des questions), où un petit modèle pourrait sinon boucler sur des appels au plan.
--| Persistance (v1) | En mémoire, session uniquement | Basé sur des fichiers | N/A |
--| Interaction utilisateur | Indirecte (via le modèle) | Indirecte (modèle charge) | Directe |
--L'outil `plan` complète les autres fonctionnalités de mini-me :
++## Dépannage
--- Utilisez les **compétences** pour des procédures complexes et réutilisables
--- Utilisez l'**outil plan** pour des demandes multi-étapes ad-hoc
--- Utilisez les **commandes slash** pour des actions utilisateur directes
-----
++- **L'agent ne crée pas de plan.** Donnez une liste numérotée, et dites « use the plan tool ». Avec un petit modèle, ajoutez le skill `plan-manager`.
++- **L'agent appelle `plan update` en boucle.** `update` change le statut d'UNE tâche ; il ne peut pas changer les tâches d'un plan. Un appel sans `status` valide reçoit une réponse qui reprend l'appel, nomme le champ manquant et donne un appel valide. Si un petit modèle le répète quand même, mettez `planTool: false` pour cet agent.
  ## Voir aussi
--- [Référence des commandes slash](../reference/slash-commands.md) - Pour les commandes utilisateur
++- [Référence des outils — `plan`](../reference/tools.md#plan) : chaque action, champ, résultat et erreur
--- [Référence des outils](../reference/tools.md) - Pour tous les outils intégrés
++- [Référence de la configuration](../reference/configuration.md) : `planTool`
--- [Aperçu des compétences](../explanation/skills.md) - Pour le système de compétences
++- [Référence des commandes](../reference/slash-commands.md) : `/new`, `/resume`
++- `demos/plans/` — une démo mesurée : un programme Go construit en quatre étapes planifiées, et un plan qui se termine sur une tâche en échec

modified docs/fr/reference/slash-commands.md +3 -3 Viewed

@@ -9,9 +9,9 @@
9	9	\| `/quit` \| Quitte le programme. \| Annule la génération, affiche `Goodbye.`, quitte. \|
10	10	\| `/abort` \| Affiche `⚠️ No generation in progress.` \| Annule la génération, affiche `🛑 Aborting...` ; les commandes déjà exécutées restent dans l'historique. \|
11	11	\| `/compact` \| Compresse l'historique immédiatement, sans tenir compte du seuil ; affiche le rapport `🗜️`, ou `🗜️ nothing to compact` quand aucun tour n'est plus ancien que `context.keepLastTurns`. \| Envoyée au modèle comme nouveau message (voir ci-dessous). \|
12		-\| `/new` \| Ramène l'historique au seul prompt système, réinitialise le détecteur de boucles et oublie le dernier compte de tokens du serveur ; affiche `🆕 new session: N message(s) forgotten` (N compte tous les messages sauf le prompt système). \| Traitée comme n'importe quel texte : annule la génération, conserve son historique partiel, puis s'exécute comme `/new` sur cet historique. \|
	12	+\| `/new` \| Ramène l'historique au seul prompt système, réinitialise le détecteur de boucles, oublie le dernier compte de tokens du serveur et les plans de la conversation ; affiche `🆕 new session: N message(s) forgotten` (N compte tous les messages sauf le prompt système). \| Traitée comme n'importe quel texte : annule la génération, conserve son historique partiel, puis s'exécute comme `/new` sur cet historique. \|
13	13	\| `/sessions` \| Liste les sessions enregistrées du projet, la plus récente d'abord : identifiant, dernière mise à jour, nombre de questions, première question, puis ` · ⚠️ N error(s)` quand la session a enregistré des erreurs ; `*` marque la session courante. Avec `sessions.enabled: false` : `⚠️ sessions are off (set sessions.enabled: true in the YAML)`. \| Envoyée au modèle comme nouveau message. \|
14		-\| `/resume [id]` \| Reprend une session enregistrée : la plus récemment mise à jour autre que la courante, en ignorant les vides, ou l'identifiant donné. Affiche `↩️ resumed <id> — N message(s), started with "<première question>"` (suivi de ` · ⚠️ N error(s)` quand la session a enregistré des erreurs) ; réinitialise le détecteur de boucles et le compte de tokens du serveur ; les tours suivants sont enregistrés dans cette session. Refusée avec une raison `⚠️` quand les sessions sont désactivées, qu'il n'y a pas de session antérieure, que l'identifiant est inconnu ou désigne la session courante. \| Envoyée au modèle comme nouveau message. \|
	14	+\| `/resume [id]` \| Reprend une session enregistrée : la plus récemment mise à jour autre que la courante, en ignorant les vides, ou l'identifiant donné. Affiche `↩️ resumed <id> — N message(s), started with "<première question>"` (suivi de ` · ⚠️ N error(s)` quand la session a enregistré des erreurs) ; réinitialise le détecteur de boucles, le compte de tokens du serveur et les plans ; les tours suivants sont enregistrés dans cette session. Refusée avec une raison `⚠️` quand les sessions sont désactivées, qu'il n'y a pas de session antérieure, que l'identifiant est inconnu ou désigne la session courante. \| Envoyée au modèle comme nouveau message. \|
15	15	\| `/help` \| Affiche les commandes de cette façade, une par ligne avec leur description, alignées. \| Envoyée au modèle comme nouveau message. \|
16	16	\| `/save <fichier> [n]` \| Écrit la dernière réponse (`n` : la n-ième depuis la fin) telle quelle dans un nouveau fichier, en créant les répertoires manquants, et affiche `💾 saved the last answer to <fichier> (N bytes)`. N'écrase jamais : un fichier existant donne `⚠️ <fichier> already exists — nothing was written; choose another name`. Un nom de fichier avec des espaces n'est pas pris en charge. Le modèle n'est pas appelé. \| Envoyée au modèle comme nouveau message. \|
17	17	\| `/memory <mots>` \| Avec `sessions.enabled` et `sessions.searchMemory` : cherche dans les sessions enregistrées du projet et affiche `🔎 N hit(s) for "<mots>" in <rép>`, puis deux lignes par résultat — `<id de session> · <date> · turn <n> · question\\|answer\\|summary` et le texte autour des mots trouvés, plus `→ <réponse>` sous une question (la réponse sur laquelle son tour s'est terminé) — ou `🔎 no hit for "<mots>" in <rép>`. La casse et les accents ne comptent pas ; les mots d'une ou deux lettres sont ignorés à côté de mots plus longs ; un résultat contient au moins un des mots ; les résultats sont classés par nombre de mots contenus, puis du plus récent au plus ancien ; 10 au plus. Les questions, les réponses et les résumés de `/compact` sont cherchés, pas les appels d'outils ; la conversation en cours est exclue. Sans mots : `⚠️ nothing to search: give one or more words`. Le modèle n'est pas appelé ; l'historique n'est pas touché. \| Envoyée au modèle comme nouveau message. \|
@@ -176,7 +176,7 @@ Deux commandes existent en mode ACP, trois avec `unloadModels` (le défaut) : `/
176	176
177	177	\| Commande \| En ACP \|
178	178	\|----------\|--------\|
179		-\| `/new` \| Comme dans le terminal : historique ramené au prompt système, compte de tokens oublié ; l'identifiant de session, le `cwd` et les autorisations « toujours autoriser » sont conservés. \|
	179	+\| `/new` \| Comme dans le terminal : historique ramené au prompt système, compte de tokens et plans de cette session oubliés (les autres sessions gardent les leurs) ; l'identifiant de session, le `cwd` et les autorisations « toujours autoriser » sont conservés. \|
180	180	\| `/compact` \| Comme dans le terminal, toujours forcée : `🗜️ compressed …`, `🗜️ nothing to compact: …` ou `[compact: failed, history kept: …]`. Attend d'abord la fin d'un tour en cours. \|
181	181	\| `/memory <mots>` \| Avec `sessions.enabled` et `sessions.searchMemory`, annoncée avec l'indication de saisie `<words>` : comme dans le terminal, dans le projet de la session (`sessions.dir` sous son `cwd`), comme texte de l'agent ; l'historique n'est pas touché. \|
182	182	\| `/docs <mots>` \| Avec `docSearch.searchDocs`, annoncée avec l'indication de saisie `<words>` : comme dans le terminal, depuis le `cwd` de la session, comme texte de l'agent ; l'historique n'est pas touché. \|

  | `/quit` | Quitte le programme. | Annule la génération, affiche `Goodbye.`, quitte. |
  | `/abort` | Affiche `⚠️ No generation in progress.` | Annule la génération, affiche `🛑 Aborting...` ; les commandes déjà exécutées restent dans l'historique. |
  | `/compact` | Compresse l'historique immédiatement, sans tenir compte du seuil ; affiche le rapport `🗜️`, ou `🗜️ nothing to compact` quand aucun tour n'est plus ancien que `context.keepLastTurns`. | Envoyée au modèle comme nouveau message (voir ci-dessous). |
--| `/new` | Ramène l'historique au seul prompt système, réinitialise le détecteur de boucles et oublie le dernier compte de tokens du serveur ; affiche `🆕 new session: N message(s) forgotten` (N compte tous les messages sauf le prompt système). | Traitée comme n'importe quel texte : annule la génération, conserve son historique partiel, puis s'exécute comme `/new` sur cet historique. |
++| `/new` | Ramène l'historique au seul prompt système, réinitialise le détecteur de boucles, oublie le dernier compte de tokens du serveur et les plans de la conversation ; affiche `🆕 new session: N message(s) forgotten` (N compte tous les messages sauf le prompt système). | Traitée comme n'importe quel texte : annule la génération, conserve son historique partiel, puis s'exécute comme `/new` sur cet historique. |
  | `/sessions` | Liste les sessions enregistrées du projet, la plus récente d'abord : identifiant, dernière mise à jour, nombre de questions, première question, puis ` · ⚠️ N error(s)` quand la session a enregistré des erreurs ; `*` marque la session courante. Avec `sessions.enabled: false` : `⚠️ sessions are off (set sessions.enabled: true in the YAML)`. | Envoyée au modèle comme nouveau message. |
--| `/resume [id]` | Reprend une session enregistrée : la plus récemment mise à jour autre que la courante, en ignorant les vides, ou l'identifiant donné. Affiche `↩️ resumed <id> — N message(s), started with "<première question>"` (suivi de ` · ⚠️ N error(s)` quand la session a enregistré des erreurs) ; réinitialise le détecteur de boucles et le compte de tokens du serveur ; les tours suivants sont enregistrés dans cette session. Refusée avec une raison `⚠️` quand les sessions sont désactivées, qu'il n'y a pas de session antérieure, que l'identifiant est inconnu ou désigne la session courante. | Envoyée au modèle comme nouveau message. |
++| `/resume [id]` | Reprend une session enregistrée : la plus récemment mise à jour autre que la courante, en ignorant les vides, ou l'identifiant donné. Affiche `↩️ resumed <id> — N message(s), started with "<première question>"` (suivi de ` · ⚠️ N error(s)` quand la session a enregistré des erreurs) ; réinitialise le détecteur de boucles, le compte de tokens du serveur et les plans ; les tours suivants sont enregistrés dans cette session. Refusée avec une raison `⚠️` quand les sessions sont désactivées, qu'il n'y a pas de session antérieure, que l'identifiant est inconnu ou désigne la session courante. | Envoyée au modèle comme nouveau message. |
  | `/help` | Affiche les commandes de cette façade, une par ligne avec leur description, alignées. | Envoyée au modèle comme nouveau message. |
  | `/save <fichier> [n]` | Écrit la dernière réponse (`n` : la n-ième depuis la fin) telle quelle dans un nouveau fichier, en créant les répertoires manquants, et affiche `💾 saved the last answer to <fichier> (N bytes)`. N'écrase jamais : un fichier existant donne `⚠️ <fichier> already exists — nothing was written; choose another name`. Un nom de fichier avec des espaces n'est pas pris en charge. Le modèle n'est pas appelé. | Envoyée au modèle comme nouveau message. |
  | `/memory <mots>` | Avec `sessions.enabled` et `sessions.searchMemory` : cherche dans les sessions enregistrées du projet et affiche `🔎 N hit(s) for "<mots>" in <rép>`, puis deux lignes par résultat — `<id de session> · <date> · turn <n> · question\|answer\|summary` et le texte autour des mots trouvés, plus `→ <réponse>` sous une question (la réponse sur laquelle son tour s'est terminé) — ou `🔎 no hit for "<mots>" in <rép>`. La casse et les accents ne comptent pas ; les mots d'une ou deux lettres sont ignorés à côté de mots plus longs ; un résultat contient au moins un des mots ; les résultats sont classés par nombre de mots contenus, puis du plus récent au plus ancien ; 10 au plus. Les questions, les réponses et les résumés de `/compact` sont cherchés, pas les appels d'outils ; la conversation en cours est exclue. Sans mots : `⚠️ nothing to search: give one or more words`. Le modèle n'est pas appelé ; l'historique n'est pas touché. | Envoyée au modèle comme nouveau message. |
  | Commande | En ACP |
  |----------|--------|
--| `/new` | Comme dans le terminal : historique ramené au prompt système, compte de tokens oublié ; l'identifiant de session, le `cwd` et les autorisations « toujours autoriser » sont conservés. |
++| `/new` | Comme dans le terminal : historique ramené au prompt système, compte de tokens et plans de cette session oubliés (les autres sessions gardent les leurs) ; l'identifiant de session, le `cwd` et les autorisations « toujours autoriser » sont conservés. |
  | `/compact` | Comme dans le terminal, toujours forcée : `🗜️ compressed …`, `🗜️ nothing to compact: …` ou `[compact: failed, history kept: …]`. Attend d'abord la fin d'un tour en cours. |
  | `/memory <mots>` | Avec `sessions.enabled` et `sessions.searchMemory`, annoncée avec l'indication de saisie `<words>` : comme dans le terminal, dans le projet de la session (`sessions.dir` sous son `cwd`), comme texte de l'agent ; l'historique n'est pas touché. |
  | `/docs <mots>` | Avec `docSearch.searchDocs`, annoncée avec l'indication de saisie `<words>` : comme dans le terminal, depuis le `cwd` de la session, comme texte de l'agent ; l'historique n'est pas touché. |

modified docs/fr/reference/tools.md +32 -0 Viewed

@@ -132,6 +132,38 @@ the closest the file comes to it — copy these lines:
132	132	│ 6 │ }
133	133	```
134	134
	135	+## `plan`
	136	+
	137	+Déclaré quand `planTool: true` (le défaut). Transforme une liste de tâches en un plan que le modèle met à jour au fil du travail. Voir [Gérer des plans et des tâches](../how-to/manage-plans-and-tasks.md).
	138	+
	139	+\| Champ \| Type \| Description \|
	140	+\|-------\|------\|-------------\|
	141	+\| `action` \| chaîne \| Obligatoire. `create`, `get`, `update`, `next`, `list`, `delete` ou `status`. \|
	142	+\| `name` \| chaîne \| Le plan. Pour `create`, facultatif : un plan sans nom s'appelle `user-request-N` (N compte les plans de la conversation). Pour les autres actions, facultatif aussi : sans lui, l'action porte sur le dernier plan créé dans la conversation (`list` l'ignore). \|
	143	+\| `tasks` \| liste de `{description, id?}` \| Pour `create` : les tâches, dans l'ordre. Une tâche sans `id` reçoit sa position (1, 2, …). Ignoré par les autres actions. \|
	144	+\| `taskId` \| entier \| Pour `update` : la tâche, à partir de 1. \|
	145	+\| `status` \| chaîne \| Pour `update` : `open`, `in_progress`, `done` ou `failed`. \|
	146	+\| `message` \| chaîne \| Pour `update` : une note gardée sur la tâche. \|
	147	+
	148	+\| Action \| Effet \| Résultat (texte JSON) \|
	149	+\|--------\|-------\|-----------------------\|
	150	+\| `create` \| Crée le plan, chaque tâche `open`. Refusé si le nom est déjà pris. \| `result` (`Plan '<nom>' created with N tasks. Current: Task 1`), `planName`, `status`, `current`, `total` ; en mode ligne, aussi le `plan` entier. \|
	151	+\| `get` \| Lit un plan. \| `result`, `plan` (nom, tâches avec id, description, statut, heures de début et de fin, note). \|
	152	+\| `update` \| Fixe le statut d'une tâche ; `in_progress` enregistre son heure de début (une fois) et en fait la tâche courante, `done` et `failed` enregistrent son heure de fin. \| `result`, `status` (la ligne de statut, ci-dessous). \|
	153	+\| `next` \| Marque la tâche courante `done` et démarre la suivante qui n'est ni faite ni en échec (`in_progress`) ; une tâche déjà terminée avec `update` est sautée. \| `result` (`Task N done. Next: Task M - <description>`, ou `Plan '<nom>' is complete! All N tasks done.`), `status`, `nextTask` s'il y en a une. \|
	154	+\| `list` \| Liste les plans existants. \| `result`, `plans` (noms, du plus ancien au plus récent). \|
	155	+\| `delete` \| Supprime un plan. \| `result`. \|
	156	+\| `status` \| Lit la ligne de statut. \| `result` (`Plan '<nom>': D/N tasks complete`), `status`. \|
	157	+
	158	+\| Aspect \| Comportement \|
	159	+\|--------\|--------------\|
	160	+\| Ligne de statut \| `[C/N] <icône> Task C: <description> (<statut>)` pour la tâche en cours de travail — la première en cours, sinon la première encore ouverte —, `[N/N] Plan finished: D done, F failed` quand toutes les tâches sont terminées et que certaines ont échoué, `[N/N] All tasks complete!` quand toutes sont faites. Icônes : ⏳ ouverte, 🔄 en cours, ✅ faite, ❌ en échec. \|
	161	+\| Erreurs \| Renvoyées comme texte, jamais comme un échec de génération : `[error: no plans found for this session]`, `[error: no plan named '<nom>' found]`, `[error: plan '<nom>' already exists]`, `[error: no task with id N in plan '<nom>']`, `[error: unknown action '<action>'. Valid actions: …]`. Un `update` sans `status` valide renvoie l'appel reçu, précise que `tasks` est ignoré par `update` quand il a été envoyé, et donne un appel qui fonctionne. \|
	162	+\| Stockage \| En mémoire, par conversation : le mode ligne et la fenêtre en tiennent une ; en ACP chaque session a la sienne, et un fil ne voit jamais les plans d'un autre. Les plans ne sont pas écrits dans le fichier de session et sont perdus à l'arrêt de mm. Une conversation qui change repart sans plan : `/new` et `/resume` dans toutes les façades, `session/load` et `session/resume` en ACP. \|
	163	+\| Permission \| `create`, `update`, `next` et `delete` sont demandés quand `yolo: false` ; un refus renvoie `[plan: command rejected by the user]`. `get`, `list` et `status` ne demandent jamais. \|
	164	+\| Affichage \| Mode ligne : `📋 plan: <action> <nom>` avant l'appel, puis `📋 <ligne de statut>` après chaque action qui en a une (`create`, `update`, `next`, `status`). Fenêtre : l'appel comme appel d'outil, et `📋 <ligne de statut>` en notice. Éditeur (ACP) : l'appel comme appel d'outil, le statut dans son résultat. La progression est affichée par mm, que le modèle la reprenne ou non. \|
	165	+\| ACP et fenêtre \| Appel d'outil de type `edit` pour `create`, `update`, `next`, `delete`, `read` pour les autres ; un résultat commençant par `[error:` est un appel en échec. \|
	166	+
135	167	## `lsp`
136	168
137	169	Déclaré quand `lsp.servers` liste au moins un serveur dont la commande est sur le `PATH` au démarrage. Un serveur listé mais introuvable est ignoré avec un `[warning: lsp server <nom> (<commande>) not found on the PATH — skipped]` ; quand aucun n'est trouvé, l'outil n'est pas déclaré et la bannière le dit.

    │ 6 │ }
  ```
++## `plan`
++
++Déclaré quand `planTool: true` (le défaut). Transforme une liste de tâches en un plan que le modèle met à jour au fil du travail. Voir [Gérer des plans et des tâches](../how-to/manage-plans-and-tasks.md).
++
++| Champ | Type | Description |
++|-------|------|-------------|
++| `action` | chaîne | Obligatoire. `create`, `get`, `update`, `next`, `list`, `delete` ou `status`. |
++| `name` | chaîne | Le plan. Pour `create`, facultatif : un plan sans nom s'appelle `user-request-N` (N compte les plans de la conversation). Pour les autres actions, facultatif aussi : sans lui, l'action porte sur le dernier plan créé dans la conversation (`list` l'ignore). |
++| `tasks` | liste de `{description, id?}` | Pour `create` : les tâches, dans l'ordre. Une tâche sans `id` reçoit sa position (1, 2, …). Ignoré par les autres actions. |
++| `taskId` | entier | Pour `update` : la tâche, à partir de 1. |
++| `status` | chaîne | Pour `update` : `open`, `in_progress`, `done` ou `failed`. |
++| `message` | chaîne | Pour `update` : une note gardée sur la tâche. |
++
++| Action | Effet | Résultat (texte JSON) |
++|--------|-------|-----------------------|
++| `create` | Crée le plan, chaque tâche `open`. Refusé si le nom est déjà pris. | `result` (`Plan '<nom>' created with N tasks. Current: Task 1`), `planName`, `status`, `current`, `total` ; en mode ligne, aussi le `plan` entier. |
++| `get` | Lit un plan. | `result`, `plan` (nom, tâches avec id, description, statut, heures de début et de fin, note). |
++| `update` | Fixe le statut d'une tâche ; `in_progress` enregistre son heure de début (une fois) et en fait la tâche courante, `done` et `failed` enregistrent son heure de fin. | `result`, `status` (la ligne de statut, ci-dessous). |
++| `next` | Marque la tâche courante `done` et démarre la suivante qui n'est ni faite ni en échec (`in_progress`) ; une tâche déjà terminée avec `update` est sautée. | `result` (`Task N done. Next: Task M - <description>`, ou `Plan '<nom>' is complete! All N tasks done.`), `status`, `nextTask` s'il y en a une. |
++| `list` | Liste les plans existants. | `result`, `plans` (noms, du plus ancien au plus récent). |
++| `delete` | Supprime un plan. | `result`. |
++| `status` | Lit la ligne de statut. | `result` (`Plan '<nom>': D/N tasks complete`), `status`. |
++
++| Aspect | Comportement |
++|--------|--------------|
++| Ligne de statut | `[C/N] <icône> Task C: <description> (<statut>)` pour la tâche en cours de travail — la première en cours, sinon la première encore ouverte —, `[N/N] Plan finished: D done, F failed` quand toutes les tâches sont terminées et que certaines ont échoué, `[N/N] All tasks complete!` quand toutes sont faites. Icônes : ⏳ ouverte, 🔄 en cours, ✅ faite, ❌ en échec. |
++| Erreurs | Renvoyées comme texte, jamais comme un échec de génération : `[error: no plans found for this session]`, `[error: no plan named '<nom>' found]`, `[error: plan '<nom>' already exists]`, `[error: no task with id N in plan '<nom>']`, `[error: unknown action '<action>'. Valid actions: …]`. Un `update` sans `status` valide renvoie l'appel reçu, précise que `tasks` est ignoré par `update` quand il a été envoyé, et donne un appel qui fonctionne. |
++| Stockage | En mémoire, par conversation : le mode ligne et la fenêtre en tiennent une ; en ACP chaque session a la sienne, et un fil ne voit jamais les plans d'un autre. Les plans ne sont pas écrits dans le fichier de session et sont perdus à l'arrêt de mm. Une conversation qui change repart sans plan : `/new` et `/resume` dans toutes les façades, `session/load` et `session/resume` en ACP. |
++| Permission | `create`, `update`, `next` et `delete` sont demandés quand `yolo: false` ; un refus renvoie `[plan: command rejected by the user]`. `get`, `list` et `status` ne demandent jamais. |
++| Affichage | Mode ligne : `📋 plan: <action> <nom>` avant l'appel, puis `📋 <ligne de statut>` après chaque action qui en a une (`create`, `update`, `next`, `status`). Fenêtre : l'appel comme appel d'outil, et `📋 <ligne de statut>` en notice. Éditeur (ACP) : l'appel comme appel d'outil, le statut dans son résultat. La progression est affichée par mm, que le modèle la reprenne ou non. |
++| ACP et fenêtre | Appel d'outil de type `edit` pour `create`, `update`, `next`, `delete`, `read` pour les autres ; un résultat commençant par `[error:` est un appel en échec. |
++
  ## `lsp`
  Déclaré quand `lsp.servers` liste au moins un serveur dont la commande est sur le `PATH` au démarrage. Un serveur listé mais introuvable est ignoré avec un `[warning: lsp server <nom> (<commande>) not found on the PATH — skipped]` ; quand aucun n'est trouvé, l'outil n'est pas déclaré et la bannière le dit.

modified internal/acp/acp.go +2 -0 Viewed

@@ -41,6 +41,7 @@ import (
41	41	history "mm/internal/session"
42	42	"mm/internal/skills"
43	43	"mm/internal/store"
	44	+ "mm/internal/tools"
44	45	"mm/internal/ui"
45	46	"mm/internal/version"
46	47
@@ -573,6 +574,7 @@ func skillCommand(input string) string {
573	574	func (f front) startNewSession(ctx context.Context, sid sdk.SessionId, s session) sdk.PromptResponse {
574	575	f.archive(sid, s)
575	576	forgotten := resetSession(s, f.system)
	577	+ tools.ClearPlans(string(sid)) // this thread's plans only (tickets 108, 109)
576	578	f.e.ForgetInputTokens()
577	579	s.usage.lastUsed = 0 // measured on the history just dropped; the totals stay, the tokens were spent
578	580	f.persist(sid, s)

  	history "mm/internal/session"
  	"mm/internal/skills"
  	"mm/internal/store"
++	"mm/internal/tools"
  	"mm/internal/ui"
  	"mm/internal/version"
  func (f *front) startNewSession(ctx context.Context, sid sdk.SessionId, s *session) sdk.PromptResponse {
  	f.archive(sid, s)
  	forgotten := resetSession(s, f.system)
++	tools.ClearPlans(string(sid)) // this thread's plans only (tickets 108, 109)
  	f.e.ForgetInputTokens()
  	s.usage.lastUsed = 0 // measured on the history just dropped; the totals stay, the tokens were spent
  	f.persist(sid, s)

modified internal/acp/persist.go +4 -0 Viewed

@@ -10,6 +10,7 @@ import (
10	10	"mm/internal/agentsmd"
11	11	history "mm/internal/session"
12	12	"mm/internal/store"
	13	+ "mm/internal/tools"
13	14
14	15	sdk "github.com/coder/acp-go-sdk"
15	16	)
@@ -81,6 +82,9 @@ func (f front) restore(method string, id sdk.SessionId, cwd string) (session,
81	82	f.mu.Lock()
82	83	f.sessions[id] = s
83	84	f.mu.Unlock()
	85	+ // Plans are not in the session file: a restored thread starts with none,
	86	+ // never with what this process held under the same id (ticket 109).
	87	+ tools.ClearPlans(string(id))
84	88	return s, nil
85	89	}
86	90

  	"mm/internal/agentsmd"
  	history "mm/internal/session"
  	"mm/internal/store"
++	"mm/internal/tools"
  	sdk "github.com/coder/acp-go-sdk"
  )
  	f.mu.Lock()
  	f.sessions[id] = s
  	f.mu.Unlock()
++	// Plans are not in the session file: a restored thread starts with none,
++	// never with what this process held under the same id (ticket 109).
++	tools.ClearPlans(string(id))
  	return s, nil
  }

added internal/acp/plans_test.go +102 -0 Viewed

new file mode 100644
@@ -0,0 +1,102 @@
	1	+package acp
	2	+
	3	+import (
	4	+ "context"
	5	+ "path/filepath"
	6	+ "strings"
	7	+ "testing"
	8	+
	9	+ "mm/internal/detector"
	10	+ "mm/internal/engine"
	11	+ "mm/internal/store"
	12	+ "mm/internal/tools"
	13	+ "mm/internal/ui"
	14	+
	15	+ "github.com/firebase/genkit/go/genkit"
	16	+)
	17	+
	18	+// threadOf stands for a turn of one ACP session, as far as the plan tool can
	19	+// tell: a sink that names its conversation and draws nothing.
	20	+type threadOf struct{ id string }
	21	+
	22	+func (t threadOf) ConversationID() string { return t.id }
	23	+func (threadOf) Text(string) {}
	24	+func (threadOf) Reasoning(string) {}
	25	+func (threadOf) ToolStart(ui.ToolEvent) {}
	26	+func (threadOf) Allow(context.Context, ui.ToolEvent) bool { return true }
	27	+func (threadOf) ToolRunning(string) {}
	28	+func (threadOf) ToolEnd(string, ui.ToolResult) {}
	29	+func (threadOf) WorkDir() string { return "" }
	30	+
	31	+// /new in one editor thread clears that thread's plans and leaves the other
	32	+// threads' alone (tickets 108, 109).
	33	+func TestNewOverACPClearsOnlyItsThreadsPlans(t *testing.T) {
	34	+ tools.ClearAllPlansForTesting()
	35	+ t.Cleanup(tools.ClearAllPlansForTesting)
	36	+ g := genkit.Init(context.Background())
	37	+ tools.PlanTool(g, detector.NewLoopDetector(10, 3))
	38	+ plan := genkit.LookupTool(g, "plan")
	39	+ run := func(thread string, input map[string]any) string {
	40	+ ui.SetActive(threadOf{id: thread})
	41	+ defer ui.SetActive(nil)
	42	+ out, err := plan.RunRaw(context.Background(), input)
	43	+ if err != nil {
	44	+ t.Fatal(err)
	45	+ }
	46	+ return out.(string)
	47	+ }
	48	+
	49	+ w := newWire(t, "sys")
	50	+ frontOf(t, w).e = &engine.Engine{}
	51	+ open := func(id int) string {
	52	+ w.send(id, "session/new", `{"cwd":"`+t.TempDir()+`","mcpServers":[]}`)
	53	+ _, resp := w.until(id)
	54	+ return resp["result"].(map[string]any)["sessionId"].(string)
	55	+ }
	56	+ a, b := open(1), open(2)
	57	+ tasks := []any{map[string]any{"description": "x"}}
	58	+ run(a, map[string]any{"action": "create", "name": "pa", "tasks": tasks})
	59	+ run(b, map[string]any{"action": "create", "name": "pb", "tasks": tasks})
	60	+
	61	+ w.send(3, "session/prompt", `{"sessionId":"`+a+`","prompt":[{"type":"text","text":"/new"}]}`)
	62	+ w.until(3)
	63	+ if out := run(a, map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
	64	+ t.Errorf("thread A after its /new: %s", out)
	65	+ }
	66	+ if out := run(b, map[string]any{"action": "list"}); !strings.Contains(out, `"pb"`) {
	67	+ t.Errorf("thread B lost its plan to A's /new: %s", out)
	68	+ }
	69	+}
	70	+
	71	+// A thread restored by session/load starts with no plan, even when this
	72	+// process held plans under the same id: plans are not in the session file.
	73	+func TestLoadSessionStartsWithNoPlan(t *testing.T) {
	74	+ withSessions(t)
	75	+ tools.ClearAllPlansForTesting()
	76	+ t.Cleanup(tools.ClearAllPlansForTesting)
	77	+ g := genkit.Init(context.Background())
	78	+ tools.PlanTool(g, detector.NewLoopDetector(10, 3))
	79	+ plan := genkit.LookupTool(g, "plan")
	80	+ const id = "20260916-100000-aaaaaaaa"
	81	+ run := func(input map[string]any) string {
	82	+ ui.SetActive(threadOf{id: id})
	83	+ defer ui.SetActive(nil)
	84	+ out, err := plan.RunRaw(context.Background(), input)
	85	+ if err != nil {
	86	+ t.Fatal(err)
	87	+ }
	88	+ return out.(string)
	89	+ }
	90	+ cwd := t.TempDir()
	91	+ if err := store.Open(filepath.Join(cwd, ".mm", "sessions")).Save(&store.Session{ID: id, Cwd: cwd, Messages: storedHistory()}); err != nil {
	92	+ t.Fatal(err)
	93	+ }
	94	+ run(map[string]any{"action": "create", "name": "stale", "tasks": []any{map[string]any{"description": "x"}}})
	95	+
	96	+ w := newWire(t, "sys")
	97	+ w.send(1, "session/load", `{"sessionId":"`+id+`","cwd":"`+cwd+`","mcpServers":[]}`)
	98	+ w.until(1)
	99	+ if out := run(map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
	100	+ t.Errorf("after session/load: %s", out)
	101	+ }
	102	+}

++package acp
++
++import (
++	"context"
++	"path/filepath"
++	"strings"
++	"testing"
++
++	"mm/internal/detector"
++	"mm/internal/engine"
++	"mm/internal/store"
++	"mm/internal/tools"
++	"mm/internal/ui"
++
++	"github.com/firebase/genkit/go/genkit"
++)
++
++// threadOf stands for a turn of one ACP session, as far as the plan tool can
++// tell: a sink that names its conversation and draws nothing.
++type threadOf struct{ id string }
++
++func (t threadOf) ConversationID() string                 { return t.id }
++func (threadOf) Text(string)                              {}
++func (threadOf) Reasoning(string)                         {}
++func (threadOf) ToolStart(ui.ToolEvent)                   {}
++func (threadOf) Allow(context.Context, ui.ToolEvent) bool { return true }
++func (threadOf) ToolRunning(string)                       {}
++func (threadOf) ToolEnd(string, ui.ToolResult)            {}
++func (threadOf) WorkDir() string                          { return "" }
++
++// /new in one editor thread clears that thread's plans and leaves the other
++// threads' alone (tickets 108, 109).
++func TestNewOverACPClearsOnlyItsThreadsPlans(t *testing.T) {
++	tools.ClearAllPlansForTesting()
++	t.Cleanup(tools.ClearAllPlansForTesting)
++	g := genkit.Init(context.Background())
++	tools.PlanTool(g, detector.NewLoopDetector(10, 3))
++	plan := genkit.LookupTool(g, "plan")
++	run := func(thread string, input map[string]any) string {
++		ui.SetActive(threadOf{id: thread})
++		defer ui.SetActive(nil)
++		out, err := plan.RunRaw(context.Background(), input)
++		if err != nil {
++			t.Fatal(err)
++		}
++		return out.(string)
++	}
++
++	w := newWire(t, "sys")
++	frontOf(t, w).e = &engine.Engine{}
++	open := func(id int) string {
++		w.send(id, "session/new", `{"cwd":"`+t.TempDir()+`","mcpServers":[]}`)
++		_, resp := w.until(id)
++		return resp["result"].(map[string]any)["sessionId"].(string)
++	}
++	a, b := open(1), open(2)
++	tasks := []any{map[string]any{"description": "x"}}
++	run(a, map[string]any{"action": "create", "name": "pa", "tasks": tasks})
++	run(b, map[string]any{"action": "create", "name": "pb", "tasks": tasks})
++
++	w.send(3, "session/prompt", `{"sessionId":"`+a+`","prompt":[{"type":"text","text":"/new"}]}`)
++	w.until(3)
++	if out := run(a, map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
++		t.Errorf("thread A after its /new: %s", out)
++	}
++	if out := run(b, map[string]any{"action": "list"}); !strings.Contains(out, `"pb"`) {
++		t.Errorf("thread B lost its plan to A's /new: %s", out)
++	}
++}
++
++// A thread restored by session/load starts with no plan, even when this
++// process held plans under the same id: plans are not in the session file.
++func TestLoadSessionStartsWithNoPlan(t *testing.T) {
++	withSessions(t)
++	tools.ClearAllPlansForTesting()
++	t.Cleanup(tools.ClearAllPlansForTesting)
++	g := genkit.Init(context.Background())
++	tools.PlanTool(g, detector.NewLoopDetector(10, 3))
++	plan := genkit.LookupTool(g, "plan")
++	const id = "20260916-100000-aaaaaaaa"
++	run := func(input map[string]any) string {
++		ui.SetActive(threadOf{id: id})
++		defer ui.SetActive(nil)
++		out, err := plan.RunRaw(context.Background(), input)
++		if err != nil {
++			t.Fatal(err)
++		}
++		return out.(string)
++	}
++	cwd := t.TempDir()
++	if err := store.Open(filepath.Join(cwd, ".mm", "sessions")).Save(&store.Session{ID: id, Cwd: cwd, Messages: storedHistory()}); err != nil {
++		t.Fatal(err)
++	}
++	run(map[string]any{"action": "create", "name": "stale", "tasks": []any{map[string]any{"description": "x"}}})
++
++	w := newWire(t, "sys")
++	w.send(1, "session/load", `{"sessionId":"`+id+`","cwd":"`+cwd+`","mcpServers":[]}`)
++	w.until(1)
++	if out := run(map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
++		t.Errorf("after session/load: %s", out)
++	}
++}

modified internal/acp/sink.go +4 -0 Viewed

@@ -53,6 +53,10 @@ func (k *sink) Reasoning(text string) {
53	53	// there, and relative file paths resolve against it (ui.Resolve).
54	54	func (k *sink) WorkDir() string { return k.s.cwd }
55	55
	56	+// ConversationID is the ACP session id: what keeps one thread's plans out of
	57	+// another's (ui.Conversation).
	58	+func (k *sink) ConversationID() string { return string(k.sid) }
	59	+
56	60	// ToolStart announces the call, status pending, before anything runs — the
57	61	// same promise as the terminal's 🛠️ line: it prints BEFORE execution, so its
58	62	// absence proves nothing ran.

  // there, and relative file paths resolve against it (ui.Resolve).
  func (k *sink) WorkDir() string { return k.s.cwd }
++// ConversationID is the ACP session id: what keeps one thread's plans out of
++// another's (ui.Conversation).
++func (k *sink) ConversationID() string { return string(k.sid) }
++
  // ToolStart announces the call, status pending, before anything runs — the
  // same promise as the terminal's 🛠️ line: it prints BEFORE execution, so its
  // absence proves nothing ran.

modified internal/agent/agent.go +3 -2 Viewed

@@ -484,8 +484,8 @@ func startNewSession(e engine.Engine, d detector.LoopDetector, system string,
484	484	forgotten := session.Forgotten(old)
485	485	e.ForgetInputTokens()
486	486	d.Reset()
487		- // Clear any active plans when starting a new session
488		- tools.ClearAllPlans()
	487	+ // Plans belong to the conversation /new just closed (ticket 109).
	488	+ tools.ClearPlans("")
489	489	dimln(fmt.Sprintf("🆕 new session: %d message(s) forgotten", forgotten))
490	490	return session.Fresh(agentsmd.System(system, ""))
491	491	}
@@ -551,6 +551,7 @@ func (l loop) historyCommand(input string, messages []ai.Message) ([]*ai.Messa
551	551	messages = resumed
552	552	l.d.Reset()
553	553	l.e.ForgetInputTokens()
	554	+ tools.ClearPlans("") // the plans in memory were the previous conversation's
554	555	fmt.Println(line)
555	556	default:
556	557	return messages, false

  	forgotten := session.Forgotten(old)
  	e.ForgetInputTokens()
  	d.Reset()
--	// Clear any active plans when starting a new session
++	// Plans belong to the conversation /new just closed (ticket 109).
--	tools.ClearAllPlans()
++	tools.ClearPlans("")
  	dimln(fmt.Sprintf("🆕 new session: %d message(s) forgotten", forgotten))
  	return session.Fresh(agentsmd.System(system, ""))
  }
  		messages = resumed
  		l.d.Reset()
  		l.e.ForgetInputTokens()
++		tools.ClearPlans("") // the plans in memory were the previous conversation's
  		fmt.Println(line)
  	default:
  		return messages, false

modified internal/tools/plan.go +118 -37 Viewed

@@ -38,6 +38,7 @@ type Plan struct {
38	38	Tasks []PlanTask `json:"tasks"`
39	39	Current int `json:"current"` // 0 = not started, 1 = first task, etc.
40	40	CreatedAt string `json:"createdAt"` // ISO timestamp
	41	+ order int // creation order in the store: CreatedAt has one-second precision
41	42	}
42	43
43	44	// PlanTask represents a single task within a plan.
@@ -55,7 +56,7 @@ type planInput struct {
55	56	// Action is required: create, get, update, next, list, delete, status
56	57	Action string `json:"action" jsonschema_description:"Required. One of: create, get, update, next, list, delete, status"`
57	58	// Name is the plan identifier (default: 'default' for unnamed plans)
58		- Name string `json:"name,omitempty" jsonschema_description:"Plan name. Defaults to 'user-request' if not provided."`
	59	+ Name string `json:"name,omitempty" jsonschema_description:"Plan name. For create, optional (a name is chosen). For the other actions, optional: without it, the most recently created plan is used."`
59	60	// Tasks is used with 'create' action to define the plan's tasks
60	61	Tasks []PlanTaskInput `json:"tasks,omitempty" jsonschema_description:"For create action: list of tasks to add"`
61	62	// TaskID identifies which task for update/next actions (1-based)
@@ -91,6 +92,7 @@ type PlanStore struct {
91	92	mu sync.RWMutex
92	93	plans map[string]map[string]*Plan // sessionID -> planName -> Plan
93	94	nextID map[string]int // sessionID -> next task ID
	95	+ made int // plans created so far, for Plan.order
94	96	}
95	97
96	98	var planStore = &PlanStore{
@@ -98,14 +100,13 @@ var planStore = &PlanStore{
98	100	nextID: make(map[string]int),
99	101	}
100	102
101		-// getSessionID extracts a session identifier from the tool context.
102		-// For simplicity in v1, we use a placeholder since the actual session ID
103		-// isn't easily accessible from tool context. In practice, plans are cleared
104		-// on /new command via ClearPlans().
	103	+// getSessionID is the conversation the plan call belongs to (ticket 108): the
	104	+// ACP session id, or "" in the line mode and the window, which hold one
	105	+// conversation at a time. It used to be a constant, so every thread an
	106	+// editor opened in one mm process saw — and could change — the others'
	107	+// plans.
105	108	func getSessionID() string {
106		- // v1: use a single session for simplicity
107		- // v2: extract from context when available
108		- return "current-session"
	109	+ return ui.ConversationID()
109	110	}
110	111
111	112	// newPlanName generates a unique plan name if none is provided.
@@ -153,11 +154,13 @@ func (s PlanStore) CreatePlan(name string, tasks []PlanTaskInput) (Plan, strin
153	154	}
154	155	}
155	156
	157	+ s.made++
156	158	plan := &Plan{
157	159	Name: name,
158	160	Tasks: planTasks,
159	161	Current: 0,
160	162	CreatedAt: time.Now().UTC().Format(time.RFC3339),
	163	+ order: s.made,
161	164	}
162	165
163	166	s.plans[sessionID][name] = plan
@@ -246,6 +249,9 @@ func (s *PlanStore) UpdateTask(name string, taskID int, status string, message s
246	249	task := &plan.Tasks[taskIndex]
247	250	task.Status = status
248	251	task.Notes = message
	252	+ if status == "in_progress" {
	253	+ plan.Current = taskIndex + 1 // so that a later `next` moves on from here
	254	+ }
249	255
250	256	// Set timestamps
251	257	now := time.Now().UTC().Format(time.RFC3339)
@@ -286,8 +292,12 @@ func (s PlanStore) NextTask(name string) (Plan, *PlanTask, string, error) {
286	292	plan.Tasks[plan.Current-1].CompletedAt = time.Now().UTC().Format(time.RFC3339)
287	293	}
288	294
289		- // Move to next
	295	+ // Move to the next task that still needs work: a task the model already
	296	+ // marked done with `update` is skipped, never started again.
290	297	plan.Current++
	298	+ for plan.Current <= len(plan.Tasks) && finished(plan.Tasks[plan.Current-1]) {
	299	+ plan.Current++
	300	+ }
291	301
292	302	// Return next task
293	303	if plan.Current <= len(plan.Tasks) {
@@ -345,28 +355,42 @@ func (s *PlanStore) GetStatus(name string) (string, string, error) {
345	355	}
346	356	}
347	357
348		- // Format status line
349		- // Current=0 means not started yet, Task 1 is next
350		- // Current>0 means we've started, Task Current is the next one
351		- if plan.Current == 0 {
352		- // Not started, show Task 1
353		- task := plan.Tasks[0]
354		- return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", 1, len(plan.Tasks), statusIcon(task.Status), 1, task.Description, task.Status),
355		- fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks)), nil
	358	+ // The line shows the task being worked on — the first in progress, else
	359	+ // the first still open — whatever the model used to get there: `update`
	360	+ // does not move plan.Current, so following Current alone showed task 1
	361	+ // for a plan whose task 1 was done and task 2 in progress (ticket 110).
	362	+ summary := fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks))
	363	+ if i := workingTask(plan); i >= 0 {
	364	+ task := plan.Tasks[i]
	365	+ return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", i+1, len(plan.Tasks), statusIcon(task.Status), i+1, task.Description, task.Status), summary, nil
356	366	}
357		-
358		- // Current > 0, show Task Current
359		- if plan.Current <= len(plan.Tasks) {
360		- task := plan.Tasks[plan.Current-1]
361		- return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", plan.Current, len(plan.Tasks), statusIcon(task.Status), plan.Current, task.Description, task.Status),
362		- fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks)), nil
	367	+ if doneCount < len(plan.Tasks) {
	368	+ failed := len(plan.Tasks) - doneCount
	369	+ return fmt.Sprintf("[%d/%d] Plan finished: %d done, %d failed", len(plan.Tasks), len(plan.Tasks), doneCount, failed), summary, nil
363	370	}
364		-
365		- // All tasks complete
366	371	return fmt.Sprintf("[%d/%d] All tasks complete!", len(plan.Tasks), len(plan.Tasks)),
367	372	fmt.Sprintf("Plan '%s': All %d tasks done ✅", name, len(plan.Tasks)), nil
368	373	}
369	374
	375	+// workingTask is the index of the task a plan is at: the first in progress,
	376	+// else the first still open; -1 when every task is done or failed.
	377	+func workingTask(plan *Plan) int {
	378	+ for i, t := range plan.Tasks {
	379	+ if t.Status == "in_progress" {
	380	+ return i
	381	+ }
	382	+ }
	383	+ for i, t := range plan.Tasks {
	384	+ if t.Status == "open" {
	385	+ return i
	386	+ }
	387	+ }
	388	+ return -1
	389	+}
	390	+
	391	+// finished reports whether a task needs no more work.
	392	+func finished(t PlanTask) bool { return t.Status == "done" \|\| t.Status == "failed" }
	393	+
370	394	// statusIcon is the display mark for a task status.
371	395	func statusIcon(status string) string {
372	396	switch status {
@@ -381,13 +405,29 @@ func statusIcon(status string) string {
381	405	}
382	406	}
383	407
384		-// ClearPlans removes all plans for the current session.
385		-// Called when /new command is issued.
386		-func (s *PlanStore) ClearPlans() {
387		- sessionID := getSessionID()
	408	+// LatestName is the name of the conversation's most recently created plan,
	409	+// "" when it has none. It is what an action without a name means: a small
	410	+// model does not keep the name `create` returned — measured on Qwen3.5 4B
	411	+// (demos/plans, 2026-10-21), every `update` came without one, failed with
	412	+// "no plan named ”", and the turn ended in runaway tool calls.
	413	+func (s *PlanStore) LatestName() string {
	414	+ conversation := getSessionID()
	415	+ s.mu.RLock()
	416	+ defer s.mu.RUnlock()
	417	+ latest, order := "", -1
	418	+ for name, p := range s.plans[conversation] {
	419	+ if p.order > order {
	420	+ latest, order = name, p.order
	421	+ }
	422	+ }
	423	+ return latest
	424	+}
	425	+
	426	+// ClearPlans removes the plans of one conversation.
	427	+func (s *PlanStore) ClearPlans(conversation string) {
388	428	s.mu.Lock()
389	429	defer s.mu.Unlock()
390		- delete(s.plans, sessionID)
	430	+ delete(s.plans, conversation)
391	431	}
392	432
393	433	// ClearAllPlans removes all plans. Used for testing.
@@ -434,16 +474,23 @@ Always show the user the current plan status using the status action.`
434	474	// Terminal/REPL: announce the call
435	475	spinner.Suspend(func() {
436	476	printMu.Lock()
437		- fmt.Printf("\n📋 plan: %s %s\n", in.Action, in.Name)
	477	+ fmt.Printf("\n📋 %s\n", planTitle(in))
438	478	printMu.Unlock()
439	479	})
440		- if planChanges(in.Action) && !confirmed(tctx, "plan", fmt.Sprintf("plan: %s %s", in.Action, in.Name)) {
	480	+ if planChanges(in.Action) && !confirmed(tctx, "plan", planTitle(in)) {
441	481	inputJSON, _ := json.Marshal(in)
442	482	return refuse(d, "plan", string(inputJSON), "[plan: command rejected by the user]")
443	483	}
444	484
445	485	spinner.Start("Running")
446	486	result := runPlanAction(in, true)
	487	+ if line := progressLine(result); line != "" {
	488	+ spinner.Suspend(func() {
	489	+ printMu.Lock()
	490	+ fmt.Println(line)
	491	+ printMu.Unlock()
	492	+ })
	493	+ }
447	494
448	495	// Record in loop detector
449	496	inputJSON, _ := json.Marshal(in)
@@ -451,6 +498,26 @@ Always show the user the current plan status using the status action.`
451	498	})
452	499	}
453	500
	501	+// planTitle names a call: `plan: update user-request-1`, `plan: list`.
	502	+func planTitle(in planInput) string {
	503	+ return strings.TrimSpace("plan: " + in.Action + " " + in.Name)
	504	+}
	505	+
	506	+// progressLine is what mm shows the user after a plan action (ticket 110):
	507	+// the plan's status line — `📋 [2/3] 🔄 Task 2: Add a license
	508	+// (in_progress)` — taken from the result, "" when the action has none (get,
	509	+// list, delete, an error). Shown by mm itself, so the progress does not
	510	+// depend on the model repeating it.
	511	+func progressLine(result string) string {
	512	+ var out struct {
	513	+ Status string `json:"status"`
	514	+ }
	515	+ if json.Unmarshal([]byte(result), &out) != nil \|\| out.Status == "" {
	516	+ return ""
	517	+ }
	518	+ return "📋 " + out.Status
	519	+}
	520	+
454	521	// planChanges reports whether a plan action changes a plan — the ones that
455	522	// ask for permission when yolo is off.
456	523	func planChanges(action string) bool {
@@ -467,7 +534,7 @@ func planEvented(ctx context.Context, s ui.Sink, d *detector.LoopDetector, in pl
467	534	ev := ui.ToolEvent{
468	535	ID: ui.NextCallID(),
469	536	Tool: "plan",
470		- Title: fmt.Sprintf("plan: %s %s", in.Action, in.Name),
	537	+ Title: planTitle(in),
471	538	Kind: eventKind,
472	539	Input: map[string]any{"action": in.Action, "name": in.Name, "taskId": in.TaskID, "status": in.Status},
473	540	}
@@ -492,16 +559,30 @@ func planEvented(ctx context.Context, s ui.Sink, d *detector.LoopDetector, in pl
492	559	failed := strings.HasPrefix(result, "[error:")
493	560
494	561	s.ToolEnd(ev.ID, ui.ToolResult{Output: result, Failed: failed})
	562	+ // The window shows the progress as a notice; an editor has none, and
	563	+ // shows the status inside the tool call's result.
	564	+ if n, ok := s.(ui.Noticer); ok {
	565	+ if line := progressLine(result); line != "" {
	566	+ n.Notice(line)
	567	+ }
	568	+ }
495	569
496	570	// Record in loop detector
497	571	inputJSON, _ := json.Marshal(in)
498	572	return record(d, "plan", string(inputJSON), result)
499	573	}
500	574
501		-// ClearAllPlans clears all plans for the current session.
502		-// Called when a new session starts (/new command).
503		-func ClearAllPlans() {
504		- planStore.ClearPlans()
	575	+// ClearPlans forgets the plans of a conversation: "" for the line mode and
	576	+// the window, the session id over ACP (ui.ConversationID). Plans are not
	577	+// saved in the session file, so every front end calls it when the
	578	+// conversation changes — /new, /resume, and over ACP session/load and
	579	+// session/resume (ticket 109): a new or restored conversation starts with
	580	+// no plan, never with the previous one's.
	581	+//
	582	+// tools.ClearPlans("") // the terminal's /new
	583	+// tools.ClearPlans(string(sid)) // an ACP session's /new
	584	+func ClearPlans(conversation string) {
	585	+ planStore.ClearPlans(conversation)
505	586	}
506	587
507	588	// ClearAllPlansForTesting exposes ClearAllPlans for tests.

  	Tasks     []PlanTask `json:"tasks"`
  	Current   int        `json:"current"`   // 0 = not started, 1 = first task, etc.
  	CreatedAt string     `json:"createdAt"` // ISO timestamp
++	order     int        // creation order in the store: CreatedAt has one-second precision
  }
  // PlanTask represents a single task within a plan.
  	// Action is required: create, get, update, next, list, delete, status
  	Action string `json:"action" jsonschema_description:"Required. One of: create, get, update, next, list, delete, status"`
  	// Name is the plan identifier (default: 'default' for unnamed plans)
--	Name string `json:"name,omitempty" jsonschema_description:"Plan name. Defaults to 'user-request' if not provided."`
++	Name string `json:"name,omitempty" jsonschema_description:"Plan name. For create, optional (a name is chosen). For the other actions, optional: without it, the most recently created plan is used."`
  	// Tasks is used with 'create' action to define the plan's tasks
  	Tasks []PlanTaskInput `json:"tasks,omitempty" jsonschema_description:"For create action: list of tasks to add"`
  	// TaskID identifies which task for update/next actions (1-based)
  	mu     sync.RWMutex
  	plans  map[string]map[string]*Plan // sessionID -> planName -> Plan
  	nextID map[string]int              // sessionID -> next task ID
++	made   int                         // plans created so far, for Plan.order
  }
  var planStore = &PlanStore{
  	nextID: make(map[string]int),
  }
--// getSessionID extracts a session identifier from the tool context.
++// getSessionID is the conversation the plan call belongs to (ticket 108): the
--// For simplicity in v1, we use a placeholder since the actual session ID
++// ACP session id, or "" in the line mode and the window, which hold one
--// isn't easily accessible from tool context. In practice, plans are cleared
++// conversation at a time. It used to be a constant, so every thread an
--// on /new command via ClearPlans().
++// editor opened in one mm process saw — and could change — the others'
++// plans.
  func getSessionID() string {
--	// v1: use a single session for simplicity
++	return ui.ConversationID()
--	// v2: extract from context when available
--	return "current-session"
  }
  // newPlanName generates a unique plan name if none is provided.
  		}
  	}
++	s.made++
  	plan := &Plan{
  		Name:      name,
  		Tasks:     planTasks,
  		Current:   0,
  		CreatedAt: time.Now().UTC().Format(time.RFC3339),
++		order:     s.made,
  	}
  	s.plans[sessionID][name] = plan
  	task := &plan.Tasks[taskIndex]
  	task.Status = status
  	task.Notes = message
++	if status == "in_progress" {
++		plan.Current = taskIndex + 1 // so that a later `next` moves on from here
++	}
  	// Set timestamps
  	now := time.Now().UTC().Format(time.RFC3339)
  		plan.Tasks[plan.Current-1].CompletedAt = time.Now().UTC().Format(time.RFC3339)
  	}
--	// Move to next
++	// Move to the next task that still needs work: a task the model already
++	// marked done with `update` is skipped, never started again.
  	plan.Current++
++	for plan.Current <= len(plan.Tasks) && finished(plan.Tasks[plan.Current-1]) {
++		plan.Current++
++	}
  	// Return next task
  	if plan.Current <= len(plan.Tasks) {
  		}
  	}
--	// Format status line
++	// The line shows the task being worked on — the first in progress, else
--	// Current=0 means not started yet, Task 1 is next
++	// the first still open — whatever the model used to get there: `update`
--	// Current>0 means we've started, Task Current is the next one
++	// does not move plan.Current, so following Current alone showed task 1
--	if plan.Current == 0 {
++	// for a plan whose task 1 was done and task 2 in progress (ticket 110).
--		// Not started, show Task 1
++	summary := fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks))
--		task := plan.Tasks[0]
++	if i := workingTask(plan); i >= 0 {
--		return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", 1, len(plan.Tasks), statusIcon(task.Status), 1, task.Description, task.Status),
++		task := plan.Tasks[i]
--			fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks)), nil
++		return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", i+1, len(plan.Tasks), statusIcon(task.Status), i+1, task.Description, task.Status), summary, nil
  	}
--
++	if doneCount < len(plan.Tasks) {
--	// Current > 0, show Task Current
++		failed := len(plan.Tasks) - doneCount
--	if plan.Current <= len(plan.Tasks) {
++		return fmt.Sprintf("[%d/%d] Plan finished: %d done, %d failed", len(plan.Tasks), len(plan.Tasks), doneCount, failed), summary, nil
--		task := plan.Tasks[plan.Current-1]
--		return fmt.Sprintf("[%d/%d] %s Task %d: %s (%s)", plan.Current, len(plan.Tasks), statusIcon(task.Status), plan.Current, task.Description, task.Status),
--			fmt.Sprintf("Plan '%s': %d/%d tasks complete", name, doneCount, len(plan.Tasks)), nil
  	}
--
--	// All tasks complete
  	return fmt.Sprintf("[%d/%d] All tasks complete!", len(plan.Tasks), len(plan.Tasks)),
  		fmt.Sprintf("Plan '%s': All %d tasks done ✅", name, len(plan.Tasks)), nil
  }
++// workingTask is the index of the task a plan is at: the first in progress,
++// else the first still open; -1 when every task is done or failed.
++func workingTask(plan *Plan) int {
++	for i, t := range plan.Tasks {
++		if t.Status == "in_progress" {
++			return i
++		}
++	}
++	for i, t := range plan.Tasks {
++		if t.Status == "open" {
++			return i
++		}
++	}
++	return -1
++}
++
++// finished reports whether a task needs no more work.
++func finished(t PlanTask) bool { return t.Status == "done" || t.Status == "failed" }
++
  // statusIcon is the display mark for a task status.
  func statusIcon(status string) string {
  	switch status {
  	}
  }
--// ClearPlans removes all plans for the current session.
++// LatestName is the name of the conversation's most recently created plan,
--// Called when /new command is issued.
++// "" when it has none. It is what an action without a name means: a small
--func (s *PlanStore) ClearPlans() {
++// model does not keep the name `create` returned — measured on Qwen3.5 4B
--	sessionID := getSessionID()
++// (demos/plans, 2026-10-21), every `update` came without one, failed with
++// "no plan named ”", and the turn ended in runaway tool calls.
++func (s *PlanStore) LatestName() string {
++	conversation := getSessionID()
++	s.mu.RLock()
++	defer s.mu.RUnlock()
++	latest, order := "", -1
++	for name, p := range s.plans[conversation] {
++		if p.order > order {
++			latest, order = name, p.order
++		}
++	}
++	return latest
++}
++
++// ClearPlans removes the plans of one conversation.
++func (s *PlanStore) ClearPlans(conversation string) {
  	s.mu.Lock()
  	defer s.mu.Unlock()
--	delete(s.plans, sessionID)
++	delete(s.plans, conversation)
  }
  // ClearAllPlans removes all plans. Used for testing.
  			// Terminal/REPL: announce the call
  			spinner.Suspend(func() {
  				printMu.Lock()
--				fmt.Printf("\n📋 plan: %s %s\n", in.Action, in.Name)
++				fmt.Printf("\n📋 %s\n", planTitle(in))
  				printMu.Unlock()
  			})
--			if planChanges(in.Action) && !confirmed(tctx, "plan", fmt.Sprintf("plan: %s %s", in.Action, in.Name)) {
++			if planChanges(in.Action) && !confirmed(tctx, "plan", planTitle(in)) {
  				inputJSON, _ := json.Marshal(in)
  				return refuse(d, "plan", string(inputJSON), "[plan: command rejected by the user]")
  			}
  			spinner.Start("Running")
  			result := runPlanAction(in, true)
++			if line := progressLine(result); line != "" {
++				spinner.Suspend(func() {
++					printMu.Lock()
++					fmt.Println(line)
++					printMu.Unlock()
++				})
++			}
  			// Record in loop detector
  			inputJSON, _ := json.Marshal(in)
  		})
  }
++// planTitle names a call: `plan: update user-request-1`, `plan: list`.
++func planTitle(in planInput) string {
++	return strings.TrimSpace("plan: " + in.Action + " " + in.Name)
++}
++
++// progressLine is what mm shows the user after a plan action (ticket 110):
++// the plan's status line — `📋 [2/3] 🔄 Task 2: Add a license
++// (in_progress)` — taken from the result, "" when the action has none (get,
++// list, delete, an error). Shown by mm itself, so the progress does not
++// depend on the model repeating it.
++func progressLine(result string) string {
++	var out struct {
++		Status string `json:"status"`
++	}
++	if json.Unmarshal([]byte(result), &out) != nil || out.Status == "" {
++		return ""
++	}
++	return "📋 " + out.Status
++}
++
  // planChanges reports whether a plan action changes a plan — the ones that
  // ask for permission when yolo is off.
  func planChanges(action string) bool {
  	ev := ui.ToolEvent{
  		ID:    ui.NextCallID(),
  		Tool:  "plan",
--		Title: fmt.Sprintf("plan: %s %s", in.Action, in.Name),
++		Title: planTitle(in),
  		Kind:  eventKind,
  		Input: map[string]any{"action": in.Action, "name": in.Name, "taskId": in.TaskID, "status": in.Status},
  	}
  	failed := strings.HasPrefix(result, "[error:")
  	s.ToolEnd(ev.ID, ui.ToolResult{Output: result, Failed: failed})
++	// The window shows the progress as a notice; an editor has none, and
++	// shows the status inside the tool call's result.
++	if n, ok := s.(ui.Noticer); ok {
++		if line := progressLine(result); line != "" {
++			n.Notice(line)
++		}
++	}
  	// Record in loop detector
  	inputJSON, _ := json.Marshal(in)
  	return record(d, "plan", string(inputJSON), result)
  }
--// ClearAllPlans clears all plans for the current session.
++// ClearPlans forgets the plans of a conversation: "" for the line mode and
--// Called when a new session starts (/new command).
++// the window, the session id over ACP (ui.ConversationID). Plans are not
--func ClearAllPlans() {
++// saved in the session file, so every front end calls it when the
--	planStore.ClearPlans()
++// conversation changes — /new, /resume, and over ACP session/load and
++// session/resume (ticket 109): a new or restored conversation starts with
++// no plan, never with the previous one's.
++//
++//	tools.ClearPlans("")           // the terminal's /new
++//	tools.ClearPlans(string(sid))  // an ACP session's /new
++func ClearPlans(conversation string) {
++	planStore.ClearPlans(conversation)
  }
  // ClearAllPlansForTesting exposes ClearAllPlans for tests.

modified internal/tools/plan_actions.go +5 -1 Viewed

@@ -25,6 +25,9 @@ func runPlanAction(in planInput, withPlan bool) string {
25	25	if in.Action == "create" {
26	26	return planCreate(in, withPlan)
27	27	}
	28	+ if in.Name == "" && in.Action != "list" {
	29	+ in.Name = planStore.LatestName()
	30	+ }
28	31	if run, ok := planActions[in.Action]; ok {
29	32	return run(in)
30	33	}
@@ -47,7 +50,8 @@ func planCreate(in planInput, withPlan bool) string {
47	50	if err != nil {
48	51	return planError(err)
49	52	}
50		- out := PlanOutput{Result: msg, PlanName: plan.Name, Current: plan.Current, Total: len(plan.Tasks)}
	53	+ statusLine, _, _ := planStore.GetStatus(plan.Name)
	54	+ out := PlanOutput{Result: msg, PlanName: plan.Name, Status: statusLine, Current: plan.Current, Total: len(plan.Tasks)}
51	55	if withPlan {
52	56	out.Plan = plan
53	57	}

  	if in.Action == "create" {
  		return planCreate(in, withPlan)
  	}
++	if in.Name == "" && in.Action != "list" {
++		in.Name = planStore.LatestName()
++	}
  	if run, ok := planActions[in.Action]; ok {
  		return run(in)
  	}
  	if err != nil {
  		return planError(err)
  	}
--	out := PlanOutput{Result: msg, PlanName: plan.Name, Current: plan.Current, Total: len(plan.Tasks)}
++	statusLine, _, _ := planStore.GetStatus(plan.Name)
++	out := PlanOutput{Result: msg, PlanName: plan.Name, Status: statusLine, Current: plan.Current, Total: len(plan.Tasks)}
  	if withPlan {
  		out.Plan = plan
  	}

modified internal/tools/plan_actions_test.go +58 -2 Viewed

@@ -2,6 +2,7 @@ package tools
2	2
3	3	import (
4	4	"encoding/json"
	5	+ "strconv"
5	6	"strings"
6	7	"testing"
7	8	)
@@ -29,8 +30,10 @@ func TestRunPlanAction(t *testing.T) {
29	30	}{
30	31	{planInput{Action: "get", Name: "p1"}, `"plan":`},
31	32	{planInput{Action: "update", Name: "p1", TaskID: 1, Status: "in_progress"}, `"status":"[1/2]`},
32		- {planInput{Action: "next", Name: "p1"}, `"nextTask":{"id":1,"description":"say \"hi\""}`},
33		- {planInput{Action: "status", Name: "p1"}, `"status":"[1/2]`},
	33	+ // next ends the task in progress (1) and starts the following one —
	34	+ // it used to start task 1 again (ticket 110).
	35	+ {planInput{Action: "next", Name: "p1"}, `"nextTask":{"id":2,"description":"B"}`},
	36	+ {planInput{Action: "status", Name: "p1"}, `"status":"[2/2] 🔄 Task 2: B (in_progress)"`},
34	37	{planInput{Action: "list"}, `"plans":["`},
35	38	{planInput{Action: "delete", Name: "p2"}, `{"result":"`},
36	39	{planInput{Action: "get", Name: "p2"}, "[error: "},
@@ -83,3 +86,56 @@ func TestPlanUpdateWithAnUnknownStatus(t *testing.T) {
83	86	t.Errorf("answer %s lacks the default example", got)
84	87	}
85	88	}
	89	+
	90	+// A description with quotes stays valid JSON in next's nextTask.
	91	+func TestPlanNextEscapesTheDescription(t *testing.T) {
	92	+ ClearAllPlansForTesting()
	93	+ t.Cleanup(ClearAllPlansForTesting)
	94	+ runPlanAction(planInput{Action: "create", Name: "q", Tasks: []PlanTaskInput{{Description: `say "hi"`}}}, true)
	95	+ if got := runPlanAction(planInput{Action: "next", Name: "q"}, true); !strings.Contains(got, `"nextTask":{"id":1,"description":"say \"hi\""}`) {
	96	+ t.Errorf("next: %s", got)
	97	+ }
	98	+}
	99	+
	100	+// The status line follows the tasks, however the model moved through them:
	101	+// marking task 1 done with update, then next, starts task 2 (it used to put
	102	+// the finished task 1 back in progress); a plan whose tasks are all over
	103	+// with failures says so.
	104	+func TestPlanStatusFollowsTheTasks(t *testing.T) {
	105	+ ClearAllPlansForTesting()
	106	+ t.Cleanup(ClearAllPlansForTesting)
	107	+ runPlanAction(planInput{Action: "create", Name: "s", Tasks: []PlanTaskInput{{Description: "one"}, {Description: "two"}}}, true)
	108	+ for _, c := range []struct {
	109	+ in planInput
	110	+ want string
	111	+ }{
	112	+ {planInput{Action: "update", Name: "s", TaskID: 1, Status: "done"}, "[2/2] ⏳ Task 2: two (open)"},
	113	+ {planInput{Action: "next", Name: "s"}, "[2/2] 🔄 Task 2: two (in_progress)"},
	114	+ {planInput{Action: "update", Name: "s", TaskID: 2, Status: "failed"}, "[2/2] Plan finished: 1 done, 1 failed"},
	115	+ {planInput{Action: "update", Name: "s", TaskID: 2, Status: "done"}, "[2/2] All tasks complete!"},
	116	+ } {
	117	+ if got := runPlanAction(c.in, true); !strings.Contains(got, `"status":"`+c.want+`"`) {
	118	+ t.Errorf("%s %d %s: %s, want status %q", c.in.Action, c.in.TaskID, c.in.Status, got, c.want)
	119	+ }
	120	+ }
	121	+}
	122	+
	123	+// An action without a name works on the conversation's latest plan — a
	124	+// small model does not keep the name create returned (demos/plans).
	125	+func TestPlanActionWithoutANameUsesTheLatestPlan(t *testing.T) {
	126	+ ClearAllPlansForTesting()
	127	+ t.Cleanup(ClearAllPlansForTesting)
	128	+ if got := runPlanAction(planInput{Action: "update", TaskID: 1, Status: "done"}, true); !strings.Contains(got, "[error: no plans found") {
	129	+ t.Errorf("no plan yet: %s", got)
	130	+ }
	131	+ for i := range 11 { // past user-request-9, where names stop sorting by number
	132	+ runPlanAction(planInput{Action: "create", Tasks: []PlanTaskInput{{Description: "t" + strconv.Itoa(i)}}}, true)
	133	+ }
	134	+ got := runPlanAction(planInput{Action: "update", TaskID: 1, Status: "in_progress"}, true)
	135	+ if !strings.Contains(got, "in plan 'user-request-11'") \|\| !strings.Contains(got, "Task 1: t10 (in_progress)") {
	136	+ t.Errorf("update without a name: %s", got)
	137	+ }
	138	+ if got := runPlanAction(planInput{Action: "status"}, true); !strings.Contains(got, "user-request-11") {
	139	+ t.Errorf("status without a name: %s", got)
	140	+ }
	141	+}

  import (
  	"encoding/json"
++	"strconv"
  	"strings"
  	"testing"
  )
  	}{
  		{planInput{Action: "get", Name: "p1"}, `"plan":`},
  		{planInput{Action: "update", Name: "p1", TaskID: 1, Status: "in_progress"}, `"status":"[1/2]`},
--		{planInput{Action: "next", Name: "p1"}, `"nextTask":{"id":1,"description":"say \"hi\""}`},
++		// next ends the task in progress (1) and starts the following one —
--		{planInput{Action: "status", Name: "p1"}, `"status":"[1/2]`},
++		// it used to start task 1 again (ticket 110).
++		{planInput{Action: "next", Name: "p1"}, `"nextTask":{"id":2,"description":"B"}`},
++		{planInput{Action: "status", Name: "p1"}, `"status":"[2/2] 🔄 Task 2: B (in_progress)"`},
  		{planInput{Action: "list"}, `"plans":["`},
  		{planInput{Action: "delete", Name: "p2"}, `{"result":"`},
  		{planInput{Action: "get", Name: "p2"}, "[error: "},
  		t.Errorf("answer %s lacks the default example", got)
  	}
  }
++
++// A description with quotes stays valid JSON in next's nextTask.
++func TestPlanNextEscapesTheDescription(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	runPlanAction(planInput{Action: "create", Name: "q", Tasks: []PlanTaskInput{{Description: `say "hi"`}}}, true)
++	if got := runPlanAction(planInput{Action: "next", Name: "q"}, true); !strings.Contains(got, `"nextTask":{"id":1,"description":"say \"hi\""}`) {
++		t.Errorf("next: %s", got)
++	}
++}
++
++// The status line follows the tasks, however the model moved through them:
++// marking task 1 done with update, then next, starts task 2 (it used to put
++// the finished task 1 back in progress); a plan whose tasks are all over
++// with failures says so.
++func TestPlanStatusFollowsTheTasks(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	runPlanAction(planInput{Action: "create", Name: "s", Tasks: []PlanTaskInput{{Description: "one"}, {Description: "two"}}}, true)
++	for _, c := range []struct {
++		in   planInput
++		want string
++	}{
++		{planInput{Action: "update", Name: "s", TaskID: 1, Status: "done"}, "[2/2] ⏳ Task 2: two (open)"},
++		{planInput{Action: "next", Name: "s"}, "[2/2] 🔄 Task 2: two (in_progress)"},
++		{planInput{Action: "update", Name: "s", TaskID: 2, Status: "failed"}, "[2/2] Plan finished: 1 done, 1 failed"},
++		{planInput{Action: "update", Name: "s", TaskID: 2, Status: "done"}, "[2/2] All tasks complete!"},
++	} {
++		if got := runPlanAction(c.in, true); !strings.Contains(got, `"status":"`+c.want+`"`) {
++			t.Errorf("%s %d %s: %s, want status %q", c.in.Action, c.in.TaskID, c.in.Status, got, c.want)
++		}
++	}
++}
++
++// An action without a name works on the conversation's latest plan — a
++// small model does not keep the name create returned (demos/plans).
++func TestPlanActionWithoutANameUsesTheLatestPlan(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	if got := runPlanAction(planInput{Action: "update", TaskID: 1, Status: "done"}, true); !strings.Contains(got, "[error: no plans found") {
++		t.Errorf("no plan yet: %s", got)
++	}
++	for i := range 11 { // past user-request-9, where names stop sorting by number
++		runPlanAction(planInput{Action: "create", Tasks: []PlanTaskInput{{Description: "t" + strconv.Itoa(i)}}}, true)
++	}
++	got := runPlanAction(planInput{Action: "update", TaskID: 1, Status: "in_progress"}, true)
++	if !strings.Contains(got, "in plan 'user-request-11'") || !strings.Contains(got, "Task 1: t10 (in_progress)") {
++		t.Errorf("update without a name: %s", got)
++	}
++	if got := runPlanAction(planInput{Action: "status"}, true); !strings.Contains(got, "user-request-11") {
++		t.Errorf("status without a name: %s", got)
++	}
++}

added internal/tools/plan_conversation_test.go +131 -0 Viewed

new file mode 100644
@@ -0,0 +1,131 @@
	1	+package tools
	2	+
	3	+import (
	4	+ "context"
	5	+ "io"
	6	+ "os"
	7	+ "strings"
	8	+ "testing"
	9	+
	10	+ "mm/internal/detector"
	11	+ "mm/internal/ui"
	12	+
	13	+ "github.com/firebase/genkit/go/genkit"
	14	+)
	15	+
	16	+// threadSink is an ACP-like sink: one conversation among several, with no
	17	+// notices (an editor shows the status inside the tool call's result).
	18	+type threadSink struct {
	19	+ recordingSink
	20	+ id string
	21	+}
	22	+
	23	+func (k *threadSink) ConversationID() string { return k.id }
	24	+
	25	+// windowSink is a window-like sink: one conversation, and notices.
	26	+type windowSink struct {
	27	+ recordingSink
	28	+ notices []string
	29	+}
	30	+
	31	+func (k *windowSink) Notice(text string) { k.notices = append(k.notices, text) }
	32	+
	33	+func runPlan(t testing.T, g genkit.Genkit, sink ui.Sink, input map[string]any) string {
	34	+ t.Helper()
	35	+ ui.SetActive(sink)
	36	+ defer ui.SetActive(nil)
	37	+ out, err := genkit.LookupTool(g, "plan").RunRaw(context.Background(), input)
	38	+ if err != nil {
	39	+ t.Fatal(err)
	40	+ }
	41	+ return out.(string)
	42	+}
	43	+
	44	+// Ticket 108: two editor threads served by one process keep their own plans
	45	+// — names, lists and reads never cross — and clearing one thread's plans
	46	+// leaves the other's.
	47	+func TestPlansArePerConversation(t *testing.T) {
	48	+ ClearAllPlansForTesting()
	49	+ t.Cleanup(ClearAllPlansForTesting)
	50	+ g := genkit.Init(context.Background())
	51	+ PlanTool(g, detector.NewLoopDetector(10, 3))
	52	+ a, b := &threadSink{id: "thread-a"}, &threadSink{id: "thread-b"}
	53	+ tasks := []any{map[string]any{"description": "one"}, map[string]any{"description": "two"}}
	54	+
	55	+ if out := runPlan(t, g, a, map[string]any{"action": "create", "tasks": tasks}); !strings.Contains(out, `"planName":"user-request-1"`) {
	56	+ t.Fatalf("thread A create: %s", out)
	57	+ }
	58	+ if out := runPlan(t, g, b, map[string]any{"action": "create", "tasks": tasks}); !strings.Contains(out, `"planName":"user-request-1"`) {
	59	+ t.Errorf("thread B's first plan is not user-request-1 — the names count A's plans: %s", out)
	60	+ }
	61	+ runPlan(t, g, b, map[string]any{"action": "create", "name": "only-in-b", "tasks": tasks})
	62	+ if out := runPlan(t, g, a, map[string]any{"action": "list"}); strings.Contains(out, "only-in-b") \|\| !strings.Contains(out, "Found 1 plan(s)") {
	63	+ t.Errorf("thread A lists B's plans: %s", out)
	64	+ }
	65	+ if out := runPlan(t, g, a, map[string]any{"action": "get", "name": "only-in-b"}); !strings.Contains(out, "no plan named 'only-in-b' found") {
	66	+ t.Errorf("thread A reads B's plan: %s", out)
	67	+ }
	68	+
	69	+ ClearPlans("thread-a")
	70	+ if out := runPlan(t, g, a, map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
	71	+ t.Errorf("thread A after its /new: %s", out)
	72	+ }
	73	+ if out := runPlan(t, g, b, map[string]any{"action": "list"}); !strings.Contains(out, "Found 2 plan(s)") {
	74	+ t.Errorf("clearing A touched B: %s", out)
	75	+ }
	76	+}
	77	+
	78	+// Ticket 110: mm shows the plan's progress itself — the status line, after
	79	+// every action that has one, create included — as a notice in the window.
	80	+// An editor's sink has no notices: nothing extra is sent there.
	81	+func TestPlanProgressIsShownByMm(t *testing.T) {
	82	+ ClearAllPlansForTesting()
	83	+ t.Cleanup(ClearAllPlansForTesting)
	84	+ g := genkit.Init(context.Background())
	85	+ PlanTool(g, detector.NewLoopDetector(10, 3))
	86	+ w := &windowSink{}
	87	+ tasks := []any{map[string]any{"description": "Create a README"}, map[string]any{"description": "Add a license"}}
	88	+
	89	+ runPlan(t, g, w, map[string]any{"action": "create", "name": "p", "tasks": tasks})
	90	+ runPlan(t, g, w, map[string]any{"action": "next", "name": "p"})
	91	+ runPlan(t, g, w, map[string]any{"action": "next", "name": "p"})
	92	+ runPlan(t, g, w, map[string]any{"action": "list"})
	93	+ want := []string{
	94	+ "📋 [1/2] ⏳ Task 1: Create a README (open)",
	95	+ "📋 [1/2] 🔄 Task 1: Create a README (in_progress)", // next starts task 1
	96	+ "📋 [2/2] 🔄 Task 2: Add a license (in_progress)", // next: 1 done, 2 started
	97	+ }
	98	+ if strings.Join(w.notices, "\n") != strings.Join(want, "\n") {
	99	+ t.Errorf("notices:\n%s\nwant:\n%s", strings.Join(w.notices, "\n"), strings.Join(want, "\n"))
	100	+ }
	101	+ if got := progressLine(`[error: no plan named 'x' found]`); got != "" {
	102	+ t.Errorf("an error gave a progress line: %q", got)
	103	+ }
	104	+}
	105	+
	106	+// In the line mode (no sink) the progress line is printed after the call
	107	+// line, on stdout.
	108	+func TestPlanProgressLineInTheLineMode(t *testing.T) {
	109	+ ClearAllPlansForTesting()
	110	+ t.Cleanup(ClearAllPlansForTesting)
	111	+ yoloOn(t)
	112	+ g := genkit.Init(context.Background())
	113	+ PlanTool(g, detector.NewLoopDetector(10, 3))
	114	+
	115	+ r, w, err := os.Pipe()
	116	+ if err != nil {
	117	+ t.Fatal(err)
	118	+ }
	119	+ stdout := os.Stdout
	120	+ os.Stdout = w
	121	+ _, err = genkit.LookupTool(g, "plan").RunRaw(context.Background(), map[string]any{"action": "create", "name": "p", "tasks": []any{map[string]any{"description": "Write the README"}}})
	122	+ os.Stdout = stdout
	123	+ _ = w.Close()
	124	+ if err != nil {
	125	+ t.Fatal(err)
	126	+ }
	127	+ printed, _ := io.ReadAll(r)
	128	+ if got := string(printed); !strings.Contains(got, "📋 plan: create p\n") \|\| !strings.Contains(got, "📋 [1/1] ⏳ Task 1: Write the README (open)\n") {
	129	+ t.Errorf("printed:\n%s", got)
	130	+ }
	131	+}

++package tools
++
++import (
++	"context"
++	"io"
++	"os"
++	"strings"
++	"testing"
++
++	"mm/internal/detector"
++	"mm/internal/ui"
++
++	"github.com/firebase/genkit/go/genkit"
++)
++
++// threadSink is an ACP-like sink: one conversation among several, with no
++// notices (an editor shows the status inside the tool call's result).
++type threadSink struct {
++	recordingSink
++	id string
++}
++
++func (k *threadSink) ConversationID() string { return k.id }
++
++// windowSink is a window-like sink: one conversation, and notices.
++type windowSink struct {
++	recordingSink
++	notices []string
++}
++
++func (k *windowSink) Notice(text string) { k.notices = append(k.notices, text) }
++
++func runPlan(t *testing.T, g *genkit.Genkit, sink ui.Sink, input map[string]any) string {
++	t.Helper()
++	ui.SetActive(sink)
++	defer ui.SetActive(nil)
++	out, err := genkit.LookupTool(g, "plan").RunRaw(context.Background(), input)
++	if err != nil {
++		t.Fatal(err)
++	}
++	return out.(string)
++}
++
++// Ticket 108: two editor threads served by one process keep their own plans
++// — names, lists and reads never cross — and clearing one thread's plans
++// leaves the other's.
++func TestPlansArePerConversation(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	g := genkit.Init(context.Background())
++	PlanTool(g, detector.NewLoopDetector(10, 3))
++	a, b := &threadSink{id: "thread-a"}, &threadSink{id: "thread-b"}
++	tasks := []any{map[string]any{"description": "one"}, map[string]any{"description": "two"}}
++
++	if out := runPlan(t, g, a, map[string]any{"action": "create", "tasks": tasks}); !strings.Contains(out, `"planName":"user-request-1"`) {
++		t.Fatalf("thread A create: %s", out)
++	}
++	if out := runPlan(t, g, b, map[string]any{"action": "create", "tasks": tasks}); !strings.Contains(out, `"planName":"user-request-1"`) {
++		t.Errorf("thread B's first plan is not user-request-1 — the names count A's plans: %s", out)
++	}
++	runPlan(t, g, b, map[string]any{"action": "create", "name": "only-in-b", "tasks": tasks})
++	if out := runPlan(t, g, a, map[string]any{"action": "list"}); strings.Contains(out, "only-in-b") || !strings.Contains(out, "Found 1 plan(s)") {
++		t.Errorf("thread A lists B's plans: %s", out)
++	}
++	if out := runPlan(t, g, a, map[string]any{"action": "get", "name": "only-in-b"}); !strings.Contains(out, "no plan named 'only-in-b' found") {
++		t.Errorf("thread A reads B's plan: %s", out)
++	}
++
++	ClearPlans("thread-a")
++	if out := runPlan(t, g, a, map[string]any{"action": "list"}); !strings.Contains(out, "No active plans") {
++		t.Errorf("thread A after its /new: %s", out)
++	}
++	if out := runPlan(t, g, b, map[string]any{"action": "list"}); !strings.Contains(out, "Found 2 plan(s)") {
++		t.Errorf("clearing A touched B: %s", out)
++	}
++}
++
++// Ticket 110: mm shows the plan's progress itself — the status line, after
++// every action that has one, create included — as a notice in the window.
++// An editor's sink has no notices: nothing extra is sent there.
++func TestPlanProgressIsShownByMm(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	g := genkit.Init(context.Background())
++	PlanTool(g, detector.NewLoopDetector(10, 3))
++	w := &windowSink{}
++	tasks := []any{map[string]any{"description": "Create a README"}, map[string]any{"description": "Add a license"}}
++
++	runPlan(t, g, w, map[string]any{"action": "create", "name": "p", "tasks": tasks})
++	runPlan(t, g, w, map[string]any{"action": "next", "name": "p"})
++	runPlan(t, g, w, map[string]any{"action": "next", "name": "p"})
++	runPlan(t, g, w, map[string]any{"action": "list"})
++	want := []string{
++		"📋 [1/2] ⏳ Task 1: Create a README (open)",
++		"📋 [1/2] 🔄 Task 1: Create a README (in_progress)", // next starts task 1
++		"📋 [2/2] 🔄 Task 2: Add a license (in_progress)",   // next: 1 done, 2 started
++	}
++	if strings.Join(w.notices, "\n") != strings.Join(want, "\n") {
++		t.Errorf("notices:\n%s\nwant:\n%s", strings.Join(w.notices, "\n"), strings.Join(want, "\n"))
++	}
++	if got := progressLine(`[error: no plan named 'x' found]`); got != "" {
++		t.Errorf("an error gave a progress line: %q", got)
++	}
++}
++
++// In the line mode (no sink) the progress line is printed after the call
++// line, on stdout.
++func TestPlanProgressLineInTheLineMode(t *testing.T) {
++	ClearAllPlansForTesting()
++	t.Cleanup(ClearAllPlansForTesting)
++	yoloOn(t)
++	g := genkit.Init(context.Background())
++	PlanTool(g, detector.NewLoopDetector(10, 3))
++
++	r, w, err := os.Pipe()
++	if err != nil {
++		t.Fatal(err)
++	}
++	stdout := os.Stdout
++	os.Stdout = w
++	_, err = genkit.LookupTool(g, "plan").RunRaw(context.Background(), map[string]any{"action": "create", "name": "p", "tasks": []any{map[string]any{"description": "Write the README"}}})
++	os.Stdout = stdout
++	_ = w.Close()
++	if err != nil {
++		t.Fatal(err)
++	}
++	printed, _ := io.ReadAll(r)
++	if got := string(printed); !strings.Contains(got, "📋 plan: create p\n") || !strings.Contains(got, "📋 [1/1] ⏳ Task 1: Write the README (open)\n") {
++		t.Errorf("printed:\n%s", got)
++	}
++}

modified internal/tools/plan_test.go +1 -1 Viewed

@@ -429,7 +429,7 @@ func TestPlanStore_ClearPlans(t *testing.T) {
429	429	}
430	430
431	431	// Clear plans
432		- store.ClearPlans()
	432	+ store.ClearPlans("")
433	433
434	434	// Verify plans cleared
435	435	plans, _, _ = store.ListPlans()

  	}
  	// Clear plans
--	store.ClearPlans()
++	store.ClearPlans("")
  	// Verify plans cleared
  	plans, _, _ = store.ListPlans()

modified internal/tui/commands.go +3 -0 Viewed

@@ -13,6 +13,7 @@ import (
13	13	"mm/internal/docsearch"
14	14	"mm/internal/session"
15	15	"mm/internal/store"
	16	+ "mm/internal/tools"
16	17	)
17	18
18	19	// submit is what Enter does with the box: a slash command or a question.
@@ -159,6 +160,7 @@ func (r *runner) newSession() {
159	160	r.opts.Model.ForgetInputTokens()
160	161	r.opts.Detector.Reset()
161	162	r.keep.Renew()
	163	+ tools.ClearPlans("") // plans belong to the conversation just closed (ticket 109)
162	164	r.t.Notice(fmt.Sprintf("🆕 new session: %d message(s) forgotten", forgotten))
163	165	}
164	166
@@ -171,6 +173,7 @@ func (r *runner) resume(id string) {
171	173	return
172	174	}
173	175	r.messages = msgs
	176	+ tools.ClearPlans("") // the plans in memory were the previous conversation's
174	177	r.opts.Detector.Reset()
175	178	r.opts.Model.ForgetInputTokens()
176	179	r.t.Notice(line)

  	"mm/internal/docsearch"
  	"mm/internal/session"
  	"mm/internal/store"
++	"mm/internal/tools"
  )
  // submit is what Enter does with the box: a slash command or a question.
  	r.opts.Model.ForgetInputTokens()
  	r.opts.Detector.Reset()
  	r.keep.Renew()
++	tools.ClearPlans("") // plans belong to the conversation just closed (ticket 109)
  	r.t.Notice(fmt.Sprintf("🆕 new session: %d message(s) forgotten", forgotten))
  }
  		return
  	}
  	r.messages = msgs
++	tools.ClearPlans("") // the plans in memory were the previous conversation's
  	r.opts.Detector.Reset()
  	r.opts.Model.ForgetInputTokens()
  	r.t.Notice(line)

modified internal/tui/turn_test.go +42 -0 Viewed

@@ -16,9 +16,11 @@ import (
16	16	mmhooks "mm/internal/hooks"
17	17	"mm/internal/session"
18	18	"mm/internal/store"
	19	+ "mm/internal/tools"
19	20	"mm/internal/ui"
20	21
21	22	"github.com/firebase/genkit/go/ai"
	23	+ "github.com/firebase/genkit/go/genkit"
22	24	)
23	25
24	26	// fakeModel answers with a canned text; its first call can block until
@@ -654,3 +656,43 @@ func TestRunnerDocs(t *testing.T) {
654	656	t.Errorf("notices:\n%s", n)
655	657	}
656	658	}
	659	+
	660	+// /new and /resume start the next conversation with no plan: the plans in
	661	+// memory belonged to the one just closed (ticket 109).
	662	+func TestRunnerNewAndResumeClearThePlans(t *testing.T) {
	663	+ tools.ClearAllPlansForTesting()
	664	+ t.Cleanup(tools.ClearAllPlansForTesting)
	665	+ g := genkit.Init(context.Background())
	666	+ tools.PlanTool(g, detector.NewLoopDetector(10, 3))
	667	+ plan := genkit.LookupTool(g, "plan")
	668	+ list := func() string {
	669	+ out, err := plan.RunRaw(context.Background(), map[string]any{"action": "list"})
	670	+ if err != nil {
	671	+ t.Fatal(err)
	672	+ }
	673	+ return out.(string)
	674	+ }
	675	+ create := func() {
	676	+ if _, err := plan.RunRaw(context.Background(), map[string]any{"action": "create", "name": "p", "tasks": []any{map[string]any{"description": "x"}}}); err != nil {
	677	+ t.Fatal(err)
	678	+ }
	679	+ }
	680	+
	681	+ r, _ := newTestRunner(t, &fakeModel{})
	682	+ create()
	683	+ r.newSession()
	684	+ if out := list(); !strings.Contains(out, "No active plans") {
	685	+ t.Errorf("after /new: %s", out)
	686	+ }
	687	+
	688	+ config.Cfg.Sessions = config.SessionsConfig{Enabled: true, Dir: ".mm/sessions"}
	689	+ r.keep = store.NewKeeper(r.opts.Cwd)
	690	+ if err := store.Open(r.keep.Dir()).Save(&store.Session{ID: "20260101-000000-aaaaaaaa", Messages: []*ai.Message{ai.NewUserTextMessage("old question")}}); err != nil {
	691	+ t.Fatal(err)
	692	+ }
	693	+ create()
	694	+ r.resume("20260101-000000-aaaaaaaa")
	695	+ if out := list(); !strings.Contains(out, "No active plans") {
	696	+ t.Errorf("after /resume: %s", out)
	697	+ }
	698	+}

  	mmhooks "mm/internal/hooks"
  	"mm/internal/session"
  	"mm/internal/store"
++	"mm/internal/tools"
  	"mm/internal/ui"
  	"github.com/firebase/genkit/go/ai"
++	"github.com/firebase/genkit/go/genkit"
  )
  // fakeModel answers with a canned text; its first call can block until
  		t.Errorf("notices:\n%s", n)
  	}
  }
++
++// /new and /resume start the next conversation with no plan: the plans in
++// memory belonged to the one just closed (ticket 109).
++func TestRunnerNewAndResumeClearThePlans(t *testing.T) {
++	tools.ClearAllPlansForTesting()
++	t.Cleanup(tools.ClearAllPlansForTesting)
++	g := genkit.Init(context.Background())
++	tools.PlanTool(g, detector.NewLoopDetector(10, 3))
++	plan := genkit.LookupTool(g, "plan")
++	list := func() string {
++		out, err := plan.RunRaw(context.Background(), map[string]any{"action": "list"})
++		if err != nil {
++			t.Fatal(err)
++		}
++		return out.(string)
++	}
++	create := func() {
++		if _, err := plan.RunRaw(context.Background(), map[string]any{"action": "create", "name": "p", "tasks": []any{map[string]any{"description": "x"}}}); err != nil {
++			t.Fatal(err)
++		}
++	}
++
++	r, _ := newTestRunner(t, &fakeModel{})
++	create()
++	r.newSession()
++	if out := list(); !strings.Contains(out, "No active plans") {
++		t.Errorf("after /new: %s", out)
++	}
++
++	config.Cfg.Sessions = config.SessionsConfig{Enabled: true, Dir: ".mm/sessions"}
++	r.keep = store.NewKeeper(r.opts.Cwd)
++	if err := store.Open(r.keep.Dir()).Save(&store.Session{ID: "20260101-000000-aaaaaaaa", Messages: []*ai.Message{ai.NewUserTextMessage("old question")}}); err != nil {
++		t.Fatal(err)
++	}
++	create()
++	r.resume("20260101-000000-aaaaaaaa")
++	if out := list(); !strings.Contains(out, "No active plans") {
++		t.Errorf("after /resume: %s", out)
++	}
++}

modified internal/ui/ui.go +20 -0 Viewed

@@ -114,6 +114,26 @@ type Elicitor interface {
114	114	Elicit(ctx context.Context, ev ToolEvent, f elicit.Form) elicit.Result
115	115	}
116	116
	117	+// Conversation is the optional side of a Sink that serves one conversation
	118	+// among several in the same process — ACP, where one mm serves every thread
	119	+// the editor opens. A tool that keeps state per conversation (the plan tool)
	120	+// reads it through ConversationID.
	121	+type Conversation interface {
	122	+ ConversationID() string
	123	+}
	124	+
	125	+// ConversationID names the conversation the turn in progress belongs to: the
	126	+// active sink's Conversation, or "" for the front ends that hold one
	127	+// conversation at a time (the line mode, the window).
	128	+//
	129	+// key := ui.ConversationID() // "20261021-…" over ACP, "" in a terminal
	130	+func ConversationID() string {
	131	+ if c, ok := ActiveSink().(Conversation); ok {
	132	+ return c.ConversationID()
	133	+ }
	134	+ return ""
	135	+}
	136	+
117	137	// Noticer is the optional side of a Sink that shows a one-line notice — a
118	138	// remark about the turn that is neither the answer nor a tool call, such as
119	139	// a hook that failed. The window has one (a dim transcript line); without

  	Elicit(ctx context.Context, ev ToolEvent, f elicit.Form) elicit.Result
  }
++// Conversation is the optional side of a Sink that serves one conversation
++// among several in the same process — ACP, where one mm serves every thread
++// the editor opens. A tool that keeps state per conversation (the plan tool)
++// reads it through ConversationID.
++type Conversation interface {
++	ConversationID() string
++}
++
++// ConversationID names the conversation the turn in progress belongs to: the
++// active sink's Conversation, or "" for the front ends that hold one
++// conversation at a time (the line mode, the window).
++//
++//	key := ui.ConversationID() // "20261021-…" over ACP, "" in a terminal
++func ConversationID() string {
++	if c, ok := ActiveSink().(Conversation); ok {
++		return c.ConversationID()
++	}
++	return ""
++}
++
  // Noticer is the optional side of a Sink that shows a one-line notice — a
  // remark about the turn that is neither the answer nor a tool call, such as
  // a hook that failed. The window has one (a dim transcript line); without

modified internal/version/about.go +4 -2 Viewed

@@ -26,8 +26,10 @@ func About() string {
26	26	b.WriteString(" mm [config.yaml] a terminal REPL, line by line\n")
27	27	b.WriteString(" mm -tui [config.yaml] one terminal window: conversation, input box, status bar\n")
28	28	b.WriteString(" mm -acp the Agent Client Protocol on stdio, for Zed and other editors\n\n")
29		- b.WriteString("Tools the model can call: bash, read_skill, read_file, write_file, edit_file, lsp.\n")
30		- b.WriteString("Model servers: Docker Model Runner, llama.cpp — anything speaking the OpenAI API.\n\n")
	29	+ b.WriteString("Built-in tools: bash, read_file, write_file, edit_file, read_skill, plan, ask_user, unload_model;\n")
	30	+ b.WriteString(" when the config turns them on: lsp, fetch_content, ask_agent, save_answer, search_memory, search_docs.\n")
	31	+ b.WriteString(" Plus the agent file's own tools (tools:) and MCP servers' tools.\n")
	32	+ b.WriteString("Model servers: Docker Model Runner, llmman, Ollama, llama.cpp — anything speaking the OpenAI API.\n\n")
31	33	fmt.Fprintf(&b, "%s\n", Home)
32	34	b.WriteString("Docs: docs/en (English), docs/fr (français) · mm -h for the flags")
33	35	return b.String()

  	b.WriteString("  mm [config.yaml]         a terminal REPL, line by line\n")
  	b.WriteString("  mm -tui [config.yaml]    one terminal window: conversation, input box, status bar\n")
  	b.WriteString("  mm -acp                  the Agent Client Protocol on stdio, for Zed and other editors\n\n")
--	b.WriteString("Tools the model can call: bash, read_skill, read_file, write_file, edit_file, lsp.\n")
++	b.WriteString("Built-in tools: bash, read_file, write_file, edit_file, read_skill, plan, ask_user, unload_model;\n")
--	b.WriteString("Model servers: Docker Model Runner, llama.cpp — anything speaking the OpenAI API.\n\n")
++	b.WriteString("  when the config turns them on: lsp, fetch_content, ask_agent, save_answer, search_memory, search_docs.\n")
++	b.WriteString("  Plus the agent file's own tools (tools:) and MCP servers' tools.\n")
++	b.WriteString("Model servers: Docker Model Runner, llmman, Ollama, llama.cpp — anything speaking the OpenAI API.\n\n")
  	fmt.Fprintf(&b, "%s\n", Home)
  	b.WriteString("Docs: docs/en (English), docs/fr (français) · mm -h for the flags")
  	return b.String()

modified release-notes/v0.8.0.en.md +1 -0 Viewed

@@ -47,6 +47,7 @@ Search what was said before and what is written down: recall your earlier questi
47	47	## Fixed
48	48
49	49	- Line mode: only the first of several piped lines was read (#103). `printf '/sessions\n/help\n' \| mm agent.yaml` or `mm agent.yaml < commands.txt` ran the first line and silently dropped the others. All of them now run, and a last line without a newline is no longer lost. Typing at the keyboard was not affected.
	50	+- `mm --about` listed only six tools and two model servers. It now lists every built-in tool — the ones on by default and the ones the configuration turns on — and the four servers (Docker Model Runner, llmman, Ollama, llama.cpp). A test now fails if a new built-in tool is missing from it.
50	51
51	52	## Demos
52	53

  ## Fixed
  - **Line mode: only the first of several piped lines was read** (#103). `printf '/sessions\n/help\n' | mm agent.yaml` or `mm agent.yaml < commands.txt` ran the first line and silently dropped the others. All of them now run, and a last line without a newline is no longer lost. Typing at the keyboard was not affected.
++- **`mm --about` listed only six tools and two model servers.** It now lists every built-in tool — the ones on by default and the ones the configuration turns on — and the four servers (Docker Model Runner, llmman, Ollama, llama.cpp). A test now fails if a new built-in tool is missing from it.
  ## Demos

modified release-notes/v0.8.0.fr.md +1 -0 Viewed

@@ -47,6 +47,7 @@ Retrouver ce qui a été dit et ce qui est écrit : rappeler ses questions préc
47	47	## Corrigé
48	48
49	49	- Mode ligne : seule la première de plusieurs lignes reçues par un pipe était lue (#103). `printf '/sessions\n/help\n' \| mm agent.yaml` ou `mm agent.yaml < commandes.txt` exécutait la première ligne et ignorait les autres sans rien dire. Elles s'exécutent maintenant toutes, et une dernière ligne sans retour à la ligne n'est plus perdue. La saisie au clavier n'était pas touchée.
	50	+- `mm --about` ne listait que six outils et deux serveurs de modèles. Il liste maintenant tous les outils intégrés — ceux actifs par défaut et ceux que la configuration active — et les quatre serveurs (Docker Model Runner, llmman, Ollama, llama.cpp). Un test échoue désormais si un nouvel outil intégré y manque.
50	51
51	52	## Démos
52	53

  ## Corrigé
  - **Mode ligne : seule la première de plusieurs lignes reçues par un pipe était lue** (#103). `printf '/sessions\n/help\n' | mm agent.yaml` ou `mm agent.yaml < commandes.txt` exécutait la première ligne et ignorait les autres sans rien dire. Elles s'exécutent maintenant toutes, et une dernière ligne sans retour à la ligne n'est plus perdue. La saisie au clavier n'était pas touchée.
++- **`mm --about` ne listait que six outils et deux serveurs de modèles.** Il liste maintenant tous les outils intégrés — ceux actifs par défaut et ceux que la configuration active — et les quatre serveurs (Docker Model Runner, llmman, Ollama, llama.cpp). Un test échoue désormais si un nouvel outil intégré y manque.
  ## Démos

added tools_reference_test.go +26 -0 Viewed

new file mode 100644
@@ -0,0 +1,26 @@
	1	+package main
	2	+
	3	+import (
	4	+ "os"
	5	+ "strings"
	6	+ "testing"
	7	+
	8	+ "mm/internal/config"
	9	+)
	10	+
	11	+// Every built-in tool has its own section in the tools reference, in both
	12	+// languages — `plan` went without one for months, documented only in a
	13	+// how-to.
	14	+func TestEveryBuiltinToolHasAReferenceSection(t *testing.T) {
	15	+ for _, page := range []string{"docs/en/reference/tools.md", "docs/fr/reference/tools.md"} {
	16	+ data, err := os.ReadFile(page)
	17	+ if err != nil {
	18	+ t.Fatal(err)
	19	+ }
	20	+ for _, name := range config.BuiltinToolNames {
	21	+ if !strings.Contains(string(data), "\n## `"+name+"`\n") {
	22	+ t.Errorf("%s has no \"## `%s`\" section", page, name)
	23	+ }
	24	+ }
	25	+ }
	26	+}

++package main
++
++import (
++	"os"
++	"strings"
++	"testing"
++
++	"mm/internal/config"
++)
++
++// Every built-in tool has its own section in the tools reference, in both
++// languages — `plan` went without one for months, documented only in a
++// how-to.
++func TestEveryBuiltinToolHasAReferenceSection(t *testing.T) {
++	for _, page := range []string{"docs/en/reference/tools.md", "docs/fr/reference/tools.md"} {
++		data, err := os.ReadFile(page)
++		if err != nil {
++			t.Fatal(err)
++		}
++		for _, name := range config.BuiltinToolNames {
++			if !strings.Contains(string(data), "\n## `"+name+"`\n") {
++				t.Errorf("%s has no \"## `%s`\" section", page, name)
++			}
++		}
++	}
++}

Merge fix/plan-mode into main