feat(titles): pull on fear, desire and curiosity triggers
Add a fear lever (mistakes, inverted formats, vague danger) and tag each title with the triggers it uses. The top 3 now opposes triggers so the A/B test shows which drive moves the audience.
50c9d0c parent: c9cb501 modified
backend/studio/modules/projects/prompts.py +2 -1 | @@ -114,7 +114,8 @@ TITLES_SCHEMA = { | ||
| 114 | 114 | "pattern": {"type": "string", |
| 115 | 115 | "description": "Le motif ou la formule du skill qu'il applique, en deux ou " |
| 116 | 116 | "trois mots."}, |
| 117 | - "why": {"type": "string", "description": "Une phrase : pourquoi ce public cliquerait."}, | |
| 117 | + "why": {"type": "string", "description": "Une phrase : le ou les déclencheurs qu'il active (peur, " | |
| 118 | + "désir, curiosité) et pourquoi ce public cliquerait."}, | |
| 118 | 119 | "top": {"type": "boolean", "description": f"Un de tes {MAX_TITLES} titres conseillés."}, |
| 119 | 120 | }, |
| 120 | 121 | }}, |
| @@ -114,7 +114,8 @@ TITLES_SCHEMA = { | |||
| 114 | "pattern": {"type": "string", | 114 | "pattern": {"type": "string", |
| 115 | "description": "Le motif ou la formule du skill qu'il applique, en deux ou " | 115 | "description": "Le motif ou la formule du skill qu'il applique, en deux ou " |
| 116 | "trois mots."}, | 116 | "trois mots."}, |
| 117 | - "why": {"type": "string", "description": "Une phrase : pourquoi ce public cliquerait."}, | 117 | + "why": {"type": "string", "description": "Une phrase : le ou les déclencheurs qu'il active (peur, " |
| 118 | + "désir, curiosité) et pourquoi ce public cliquerait."}, | ||
| 118 | "top": {"type": "boolean", "description": f"Un de tes {MAX_TITLES} titres conseillés."}, | 119 | "top": {"type": "boolean", "description": f"Un de tes {MAX_TITLES} titres conseillés."}, |
| 119 | }, | 120 | }, |
| 120 | }}, | 121 | }}, |
modified
skills/youtube-titles/SKILL.md +44 -5 | @@ -1,6 +1,6 @@ | ||
| 1 | 1 | --- |
| 2 | 2 | name: youtube-titles |
| 3 | -description: Generate 30 YouTube titles that earn clicks without lying — 20 built on three structural formulas (bold statement, how-to, time frame) with psychological levers, plus 10 free titles across proven click patterns (curiosity gap, outcome, contrarian, numbers, comparison, story, how-to). Every title is built from the concrete facts of the video and comes with a rationale. Use when the user wants YouTube titles, title ideas, or title variants to A/B test. | |
| 3 | +description: Generate 30 YouTube titles that earn clicks without lying — each one pulling on fear, desire or curiosity, ideally two of them — 20 built on three structural formulas (bold statement, how-to, time frame) with psychological levers, plus 10 free titles across proven click patterns (curiosity gap, outcome, contrarian, numbers, comparison, story, how-to, mistake). Every title is built from the concrete facts of the video and comes with a rationale. Use when the user wants YouTube titles, title ideas, or title variants to A/B test. | |
| 4 | 4 | --- |
| 5 | 5 | |
| 6 | 6 | # YouTube Title Generator |
| @@ -13,6 +13,36 @@ You generate YouTube titles that earn clicks without resorting to lies. Every ti | ||
| 13 | 13 | |
| 14 | 14 | You produce a wide range (30 titles), from safe to polarizing, so the author can see several kinds of promises before choosing. |
| 15 | 15 | |
| 16 | +## The Three Triggers | |
| 17 | + | |
| 18 | +Every click comes from one of three drives. Each title pulls on at least one; the strongest stack two, sometimes all three. | |
| 19 | + | |
| 20 | +### Fear | |
| 21 | +The viewer worries they're missing something, or already doing something wrong. | |
| 22 | +- **Start from the audience's real worry**: falling behind, wasting months on the wrong skill, the "good practice" that secretly hurts them. | |
| 23 | +- **Flip tips into mistakes**: "5 Ways to Write Better Prompts" → "5 Prompt Mistakes Wrecking Your Output". | |
| 24 | +- **Invert a proven format**: "How to Ship Faster" → "The Fastest Way to Wreck Your Codebase". | |
| 25 | +- **Tease the danger, don't name it**: "The Setting That Leaks Your API Keys" makes the viewer check; naming the setting answers the question. | |
| 26 | + | |
| 27 | +Only a fear the audience already has, and one the video actually addresses. Don't overdo it: one worry per title, no doom. | |
| 28 | + | |
| 29 | +### Desire | |
| 30 | +The title promises a transformation the viewer wants. | |
| 31 | +- **Sell the result, not the process**: the deployed app, the hours saved, the job — not the tutorial steps. | |
| 32 | +- The video must deliver the full result it promises, or the click turns into a thumbs-down. | |
| 33 | + | |
| 34 | +### Curiosity | |
| 35 | +The title opens a question only the video can answer. | |
| 36 | +- **Ask what nobody thought to ask**: "How Much Does a Claude Code Session Really Cost?" | |
| 37 | +- **Build in contrast**: a $5 setup next to a $500 one, 1 prompt next to 100. | |
| 38 | +- **Go novel**: something almost nobody would try ("I Let AI Run My Side Project for 30 Days"). | |
| 39 | +- **Use "actually" or "really"**: "Do AI Agents Actually Save Time?" implies the answer is surprising. | |
| 40 | + | |
| 41 | +### Stacking | |
| 42 | +- Desire + curiosity: "Get Your First 1,000 Users (Nobody Teaches This Part)" | |
| 43 | +- Fear + curiosity: "The Copilot Habit Slowing You Down (Without You Noticing)" | |
| 44 | +- Fear + desire: "Stop Learning New Frameworks (Learn This Instead)" | |
| 45 | + | |
| 16 | 46 | ## What You Need |
| 17 | 47 | |
| 18 | 48 | Ask the user for: |
| @@ -33,6 +63,7 @@ Before writing anything, list what the video actually contains. | ||
| 33 | 63 | | Transformation | What changes for the viewer: skills, career, income, time, way of working? | |
| 34 | 64 | | Big idea | The one claim from the video people would repeat to a friend | |
| 35 | 65 | | Audience | Who gains, and what they want or fear | |
| 66 | +| Risk | The mistake, waste or danger the video helps the viewer avoid — only if it really does | | |
| 36 | 67 | | Contrast | Where a modest effort leads to an outsized result, or where the advice goes against the grain | |
| 37 | 68 | |
| 38 | 69 | **The facts** |
| @@ -67,7 +98,7 @@ A concrete amount of time or effort, then where to put it. | ||
| 67 | 98 | - "2 Hours a Day for 6 Months (The Skills Worth Learning Now)" |
| 68 | 99 | - "From Screenshot to Working Page in 18 Minutes (Full Build)" |
| 69 | 100 | |
| 70 | -Spread these levers across the 20 titles, each in several titles: | |
| 101 | +Across the 20 titles, all three triggers appear several times each, and at least half the titles stack two. Spread these levers too, each in several titles: | |
| 71 | 102 | |
| 72 | 103 | | Lever | How | Wording | |
| 73 | 104 | | --- | --- | --- | |
| @@ -75,6 +106,7 @@ Spread these levers across the 20 titles, each in several titles: | ||
| 75 | 106 | | Transformation | A visible change in the viewer | "change your career", "you won't work the same way" | |
| 76 | 107 | | Insider knowledge | What most people miss | "what nobody tells you", "most people ignore this" | |
| 77 | 108 | | Status | Getting ahead | "ahead of 99 % of developers", "irreplaceable" | |
| 109 | +| Mistake / risk | What the viewer may be getting wrong | "the mistake killing your…", "before it's too late", "you're doing it wrong" | | |
| 78 | 110 | |
| 79 | 111 | And these contrasts, in several titles: |
| 80 | 112 | - **Small input, big output:** "2 hours a day" → "a new career", "one screenshot" → "a full app". |
| @@ -83,7 +115,7 @@ And these contrasts, in several titles: | ||
| 83 | 115 | |
| 84 | 116 | ## Step 3: 10 Free Titles |
| 85 | 117 | |
| 86 | -Ten more that leave the formulas aside: the most clickable titles you can find for this video, noticeably different from the first twenty. Draw on these patterns, at least five of them: | |
| 118 | +Ten more that leave the formulas aside: the most clickable titles you can find for this video, noticeably different from the first twenty. Draw on these patterns, at least five of them, and cover all three triggers: | |
| 87 | 119 | |
| 88 | 120 | ### 1. Curiosity Gap |
| 89 | 121 | Create an information gap the viewer needs to close. Say enough to intrigue, not enough to satisfy. |
| @@ -120,6 +152,11 @@ Direct value promise. Works best when the outcome is specific. | ||
| 120 | 152 | - "How to [Achieve Specific Result] (Step by Step)" |
| 121 | 153 | - "The Complete Guide to [X] for [Audience]" |
| 122 | 154 | |
| 155 | +### 8. Mistake / Warning | |
| 156 | +The fear trigger on its own: what the viewer is getting wrong, without saying exactly what. | |
| 157 | +- "[Number] [X] Mistakes Killing Your [Outcome]" | |
| 158 | +- "You're Using [Tool] Wrong" | |
| 159 | + | |
| 123 | 160 | For tech, tutorial and tool videos, the content type points to the strongest pattern: |
| 124 | 161 | |
| 125 | 162 | | The video is… | Lead with | |
| @@ -158,12 +195,13 @@ Present each title with: | ||
| 158 | 195 | ``` |
| 159 | 196 | **[Title]** |
| 160 | 197 | Pattern: [which formula or pattern] |
| 161 | -Why it works: [one sentence — what makes this clickable for the target audience] | |
| 198 | +Triggers: [fear, desire and/or curiosity] | |
| 199 | +Why it works: [one sentence — which trigger it pulls on and why this audience feels it] | |
| 162 | 200 | ``` |
| 163 | 201 | |
| 164 | 202 | ## After the List |
| 165 | 203 | |
| 166 | -Recommend your **top 3 picks** and explain why. Pick three titles that make three different promises (for example one statement, one how-to, one free title), so an A/B test tells the author which kind of promise their audience prefers. Consider: | |
| 204 | +Recommend your **top 3 picks** and explain why. Pick three titles led by three different triggers (one fear, one desire, one curiosity), so an A/B test tells the author which drive moves their audience. Within that, prefer titles that stack a second trigger. Consider: | |
| 167 | 205 | - Which ones have the strongest curiosity gap? |
| 168 | 206 | - Which ones most clearly signal value to the target audience? |
| 169 | 207 | - Which ones would stand out in a feed full of similar content? |
| @@ -173,6 +211,7 @@ If the user has existing videos (check CLAUDE.md or ask), factor in what title s | ||
| 173 | 211 | ## Rules |
| 174 | 212 | |
| 175 | 213 | - **High conviction, never a lie.** Be bold, take sides, be polarizing when it fits — but every title is a promise the video actually keeps. |
| 214 | +- **Real fears only.** A fear title names a worry the audience already has and the video resolves; never invent a threat to scare people into clicking. | |
| 176 | 215 | - **30 distinct titles.** No two titles that only differ by a word; no formula twice in a row. |
| 177 | 216 | - **The core idea stays.** Every title is about this video, not a more clickable neighbouring topic. |
| 178 | 217 | - Every title under 60 characters when possible (displays fully on mobile), 70 at most |
| @@ -1,6 +1,6 @@ | |||
| 1 | --- | 1 | --- |
| 2 | name: youtube-titles | 2 | name: youtube-titles |
| 3 | -description: Generate 30 YouTube titles that earn clicks without lying — 20 built on three structural formulas (bold statement, how-to, time frame) with psychological levers, plus 10 free titles across proven click patterns (curiosity gap, outcome, contrarian, numbers, comparison, story, how-to). Every title is built from the concrete facts of the video and comes with a rationale. Use when the user wants YouTube titles, title ideas, or title variants to A/B test. | 3 | +description: Generate 30 YouTube titles that earn clicks without lying — each one pulling on fear, desire or curiosity, ideally two of them — 20 built on three structural formulas (bold statement, how-to, time frame) with psychological levers, plus 10 free titles across proven click patterns (curiosity gap, outcome, contrarian, numbers, comparison, story, how-to, mistake). Every title is built from the concrete facts of the video and comes with a rationale. Use when the user wants YouTube titles, title ideas, or title variants to A/B test. |
| 4 | --- | 4 | --- |
| 5 | 5 | ||
| 6 | # YouTube Title Generator | 6 | # YouTube Title Generator |
| @@ -13,6 +13,36 @@ You generate YouTube titles that earn clicks without resorting to lies. Every ti | |||
| 13 | 13 | ||
| 14 | You produce a wide range (30 titles), from safe to polarizing, so the author can see several kinds of promises before choosing. | 14 | You produce a wide range (30 titles), from safe to polarizing, so the author can see several kinds of promises before choosing. |
| 15 | 15 | ||
| 16 | +## The Three Triggers | ||
| 17 | + | ||
| 18 | +Every click comes from one of three drives. Each title pulls on at least one; the strongest stack two, sometimes all three. | ||
| 19 | + | ||
| 20 | +### Fear | ||
| 21 | +The viewer worries they're missing something, or already doing something wrong. | ||
| 22 | +- **Start from the audience's real worry**: falling behind, wasting months on the wrong skill, the "good practice" that secretly hurts them. | ||
| 23 | +- **Flip tips into mistakes**: "5 Ways to Write Better Prompts" → "5 Prompt Mistakes Wrecking Your Output". | ||
| 24 | +- **Invert a proven format**: "How to Ship Faster" → "The Fastest Way to Wreck Your Codebase". | ||
| 25 | +- **Tease the danger, don't name it**: "The Setting That Leaks Your API Keys" makes the viewer check; naming the setting answers the question. | ||
| 26 | + | ||
| 27 | +Only a fear the audience already has, and one the video actually addresses. Don't overdo it: one worry per title, no doom. | ||
| 28 | + | ||
| 29 | +### Desire | ||
| 30 | +The title promises a transformation the viewer wants. | ||
| 31 | +- **Sell the result, not the process**: the deployed app, the hours saved, the job — not the tutorial steps. | ||
| 32 | +- The video must deliver the full result it promises, or the click turns into a thumbs-down. | ||
| 33 | + | ||
| 34 | +### Curiosity | ||
| 35 | +The title opens a question only the video can answer. | ||
| 36 | +- **Ask what nobody thought to ask**: "How Much Does a Claude Code Session Really Cost?" | ||
| 37 | +- **Build in contrast**: a $5 setup next to a $500 one, 1 prompt next to 100. | ||
| 38 | +- **Go novel**: something almost nobody would try ("I Let AI Run My Side Project for 30 Days"). | ||
| 39 | +- **Use "actually" or "really"**: "Do AI Agents Actually Save Time?" implies the answer is surprising. | ||
| 40 | + | ||
| 41 | +### Stacking | ||
| 42 | +- Desire + curiosity: "Get Your First 1,000 Users (Nobody Teaches This Part)" | ||
| 43 | +- Fear + curiosity: "The Copilot Habit Slowing You Down (Without You Noticing)" | ||
| 44 | +- Fear + desire: "Stop Learning New Frameworks (Learn This Instead)" | ||
| 45 | + | ||
| 16 | ## What You Need | 46 | ## What You Need |
| 17 | 47 | ||
| 18 | Ask the user for: | 48 | Ask the user for: |
| @@ -33,6 +63,7 @@ Before writing anything, list what the video actually contains. | |||
| 33 | | Transformation | What changes for the viewer: skills, career, income, time, way of working? | | 63 | | Transformation | What changes for the viewer: skills, career, income, time, way of working? | |
| 34 | | Big idea | The one claim from the video people would repeat to a friend | | 64 | | Big idea | The one claim from the video people would repeat to a friend | |
| 35 | | Audience | Who gains, and what they want or fear | | 65 | | Audience | Who gains, and what they want or fear | |
| 66 | +| Risk | The mistake, waste or danger the video helps the viewer avoid — only if it really does | | ||
| 36 | | Contrast | Where a modest effort leads to an outsized result, or where the advice goes against the grain | | 67 | | Contrast | Where a modest effort leads to an outsized result, or where the advice goes against the grain | |
| 37 | 68 | ||
| 38 | **The facts** | 69 | **The facts** |
| @@ -67,7 +98,7 @@ A concrete amount of time or effort, then where to put it. | |||
| 67 | - "2 Hours a Day for 6 Months (The Skills Worth Learning Now)" | 98 | - "2 Hours a Day for 6 Months (The Skills Worth Learning Now)" |
| 68 | - "From Screenshot to Working Page in 18 Minutes (Full Build)" | 99 | - "From Screenshot to Working Page in 18 Minutes (Full Build)" |
| 69 | 100 | ||
| 70 | -Spread these levers across the 20 titles, each in several titles: | 101 | +Across the 20 titles, all three triggers appear several times each, and at least half the titles stack two. Spread these levers too, each in several titles: |
| 71 | 102 | ||
| 72 | | Lever | How | Wording | | 103 | | Lever | How | Wording | |
| 73 | | --- | --- | --- | | 104 | | --- | --- | --- | |
| @@ -75,6 +106,7 @@ Spread these levers across the 20 titles, each in several titles: | |||
| 75 | | Transformation | A visible change in the viewer | "change your career", "you won't work the same way" | | 106 | | Transformation | A visible change in the viewer | "change your career", "you won't work the same way" | |
| 76 | | Insider knowledge | What most people miss | "what nobody tells you", "most people ignore this" | | 107 | | Insider knowledge | What most people miss | "what nobody tells you", "most people ignore this" | |
| 77 | | Status | Getting ahead | "ahead of 99 % of developers", "irreplaceable" | | 108 | | Status | Getting ahead | "ahead of 99 % of developers", "irreplaceable" | |
| 109 | +| Mistake / risk | What the viewer may be getting wrong | "the mistake killing your…", "before it's too late", "you're doing it wrong" | | ||
| 78 | 110 | ||
| 79 | And these contrasts, in several titles: | 111 | And these contrasts, in several titles: |
| 80 | - **Small input, big output:** "2 hours a day" → "a new career", "one screenshot" → "a full app". | 112 | - **Small input, big output:** "2 hours a day" → "a new career", "one screenshot" → "a full app". |
| @@ -83,7 +115,7 @@ And these contrasts, in several titles: | |||
| 83 | 115 | ||
| 84 | ## Step 3: 10 Free Titles | 116 | ## Step 3: 10 Free Titles |
| 85 | 117 | ||
| 86 | -Ten more that leave the formulas aside: the most clickable titles you can find for this video, noticeably different from the first twenty. Draw on these patterns, at least five of them: | 118 | +Ten more that leave the formulas aside: the most clickable titles you can find for this video, noticeably different from the first twenty. Draw on these patterns, at least five of them, and cover all three triggers: |
| 87 | 119 | ||
| 88 | ### 1. Curiosity Gap | 120 | ### 1. Curiosity Gap |
| 89 | Create an information gap the viewer needs to close. Say enough to intrigue, not enough to satisfy. | 121 | Create an information gap the viewer needs to close. Say enough to intrigue, not enough to satisfy. |
| @@ -120,6 +152,11 @@ Direct value promise. Works best when the outcome is specific. | |||
| 120 | - "How to [Achieve Specific Result] (Step by Step)" | 152 | - "How to [Achieve Specific Result] (Step by Step)" |
| 121 | - "The Complete Guide to [X] for [Audience]" | 153 | - "The Complete Guide to [X] for [Audience]" |
| 122 | 154 | ||
| 155 | +### 8. Mistake / Warning | ||
| 156 | +The fear trigger on its own: what the viewer is getting wrong, without saying exactly what. | ||
| 157 | +- "[Number] [X] Mistakes Killing Your [Outcome]" | ||
| 158 | +- "You're Using [Tool] Wrong" | ||
| 159 | + | ||
| 123 | For tech, tutorial and tool videos, the content type points to the strongest pattern: | 160 | For tech, tutorial and tool videos, the content type points to the strongest pattern: |
| 124 | 161 | ||
| 125 | | The video is… | Lead with | | 162 | | The video is… | Lead with | |
| @@ -158,12 +195,13 @@ Present each title with: | |||
| 158 | ``` | 195 | ``` |
| 159 | **[Title]** | 196 | **[Title]** |
| 160 | Pattern: [which formula or pattern] | 197 | Pattern: [which formula or pattern] |
| 161 | -Why it works: [one sentence — what makes this clickable for the target audience] | 198 | +Triggers: [fear, desire and/or curiosity] |
| 199 | +Why it works: [one sentence — which trigger it pulls on and why this audience feels it] | ||
| 162 | ``` | 200 | ``` |
| 163 | 201 | ||
| 164 | ## After the List | 202 | ## After the List |
| 165 | 203 | ||
| 166 | -Recommend your **top 3 picks** and explain why. Pick three titles that make three different promises (for example one statement, one how-to, one free title), so an A/B test tells the author which kind of promise their audience prefers. Consider: | 204 | +Recommend your **top 3 picks** and explain why. Pick three titles led by three different triggers (one fear, one desire, one curiosity), so an A/B test tells the author which drive moves their audience. Within that, prefer titles that stack a second trigger. Consider: |
| 167 | - Which ones have the strongest curiosity gap? | 205 | - Which ones have the strongest curiosity gap? |
| 168 | - Which ones most clearly signal value to the target audience? | 206 | - Which ones most clearly signal value to the target audience? |
| 169 | - Which ones would stand out in a feed full of similar content? | 207 | - Which ones would stand out in a feed full of similar content? |
| @@ -173,6 +211,7 @@ If the user has existing videos (check CLAUDE.md or ask), factor in what title s | |||
| 173 | ## Rules | 211 | ## Rules |
| 174 | 212 | ||
| 175 | - **High conviction, never a lie.** Be bold, take sides, be polarizing when it fits — but every title is a promise the video actually keeps. | 213 | - **High conviction, never a lie.** Be bold, take sides, be polarizing when it fits — but every title is a promise the video actually keeps. |
| 214 | +- **Real fears only.** A fear title names a worry the audience already has and the video resolves; never invent a threat to scare people into clicking. | ||
| 176 | - **30 distinct titles.** No two titles that only differ by a word; no formula twice in a row. | 215 | - **30 distinct titles.** No two titles that only differ by a word; no formula twice in a row. |
| 177 | - **The core idea stays.** Every title is about this video, not a more clickable neighbouring topic. | 216 | - **The core idea stays.** Every title is about this video, not a more clickable neighbouring topic. |
| 178 | - Every title under 60 characters when possible (displays fully on mobile), 70 at most | 217 | - Every title under 60 characters when possible (displays fully on mobile), 70 at most |