agent-prd
starsforks
How to install agent-prd?
davekilleen/the-vibe-pm/agent-prd
$npx -y skills add davekilleen/the-vibe-pm --skill agent-prdPurpose: Help users write PRDs that make AI agents more deterministic and successful.
SKILL.md
| 1 | # Agent-Oriented PRD Writer |
| 2 | |
| 3 | **Purpose:** Help users write PRDs that make AI agents more deterministic and successful. |
| 4 | |
| 5 | **Based on:** StrongDM's software factory pattern (observable behaviors, satisfaction metrics, holdout scenarios) |
| 6 | |
| 7 | **Works in:** Claude Code, Claude Desktop, Cursor, or any AI coding agent harness |
| 8 | |
| 9 | --- |
| 10 | |
| 11 | ## When This Skill Activates |
| 12 | |
| 13 | User says: |
| 14 | - `/agent-prd [feature description]` |
| 15 | - "Help me write an agent-oriented PRD" |
| 16 | - "Create a PRD for [feature]" |
| 17 | |
| 18 | --- |
| 19 | |
| 20 | ## Context Detection (Read First) |
| 21 | |
| 22 | Before starting the interactive wizard, detect user's context: |
| 23 | |
| 24 | ### 1. Company |
| 25 | Search vault for person pages, company pages. Look for: |
| 26 | - User's employer (in their profile or frequently mentioned) |
| 27 | - Current project context |
| 28 | |
| 29 | ### 2. Analytics Tools |
| 30 | Check for mentions of: |
| 31 | - Pendo |
| 32 | - Amplitude |
| 33 | - Mixpanel |
| 34 | - Google Analytics |
| 35 | - PostHog |
| 36 | |
| 37 | ### 3. Project Management Tools |
| 38 | Check for mentions of: |
| 39 | - Linear |
| 40 | - Jira |
| 41 | - Productboard |
| 42 | |
| 43 | ### 4. Existing PRDs |
| 44 | Search for similar PRDs in vault. Use as templates. |
| 45 | |
| 46 | **Output detected context to user:** |
| 47 | ``` |
| 48 | I see you're at [Company] and use [Analytics Tool] + [PM Tool]. |
| 49 | I'll tailor metrics suggestions based on this. |
| 50 | ``` |
| 51 | |
| 52 | --- |
| 53 | |
| 54 | ## Interactive PRD Wizard |
| 55 | |
| 56 | Guide user through 8 steps: |
| 57 | |
| 58 | ### Step 1: The Job to Be Done |
| 59 | **Ask:** "What should users be able to DO with this feature? (One sentence)" |
| 60 | |
| 61 | **Good answer:** "Users can access their terminal session from a mobile browser" |
| 62 | **Bad answer:** "Build a web UI component that renders..." |
| 63 | |
| 64 | **If user gives implementation details, redirect:** |
| 65 | "Let's focus on the user outcome first. What job does this feature do for them?" |
| 66 | |
| 67 | ### Step 2: Observable Behaviors |
| 68 | **Ask:** "Describe what an observer would SEE when this feature works. Be specific." |
| 69 | |
| 70 | **Prompt:** |
| 71 | ``` |
| 72 | Think about: |
| 73 | - What does the user do? (actions) |
| 74 | - What happens on screen? (visible outcomes) |
| 75 | - How fast? (time constraints) |
| 76 | - What changes? (state transitions) |
| 77 | ``` |
| 78 | |
| 79 | **Guide toward scenarios like:** |
| 80 | - User clicks X → Screen shows Y within 2 seconds |
| 81 | - User types message → Response appears in <500ms |
| 82 | - User closes laptop, opens 2 hrs later → Session reconnects automatically |
| 83 | |
| 84 | ### Step 3: Success Scenarios |
| 85 | **Ask:** "What are the critical scenarios to test? Include at least one error case." |
| 86 | |
| 87 | **Generate template:** |
| 88 | ``` |
| 89 | Scenario 1: [Happy Path] |
| 90 | Setup: [Starting conditions] |
| 91 | Action: [What user does] |
| 92 | Observable Outcome: [What observer sees] |
| 93 | Success Criteria: [Quantifiable target] |
| 94 | |
| 95 | Scenario 2: [Error Case] |
| 96 | Setup: [How to trigger error] |
| 97 | Action: [What user does] |
| 98 | Observable Outcome: [How system handles it gracefully] |
| 99 | Success Criteria: [Recovery success rate] |
| 100 | ``` |
| 101 | |
| 102 | ### Step 4: Metrics Strategy |
| 103 | **Context-aware suggestions based on detected analytics tool:** |
| 104 | |
| 105 | If analytics tool detected (e.g. Pendo, Amplitude, Mixpanel): |
| 106 | ``` |
| 107 | For this feature, I recommend tracking: |
| 108 | |
| 109 | 📊 Events: |
| 110 | - [feature]_started (when user begins) |
| 111 | - [feature]_completed (successful outcome) |
| 112 | - [feature]_error (failure modes) |
| 113 | |
| 114 | 📈 Metrics: |
| 115 | - Adoption rate (% of users who try feature) |
| 116 | - Completion rate (% who succeed) |
| 117 | - Time-to-complete (median duration) |
| 118 | |
| 119 | 🎯 Success Targets: |
| 120 | Based on industry benchmarks: |
| 121 | - [X]% adoption within 30 days |
| 122 | - [Y]% completion rate |
| 123 | - <[Z] seconds time-to-complete |
| 124 | |
| 125 | Do these metrics align with your goals? |
| 126 | ``` |
| 127 | |
| 128 | If **no analytics tool** detected: |
| 129 | ``` |
| 130 | What analytics tool do you use? |
| 131 | - Pendo |
| 132 | - Amplitude |
| 133 | - Mixpanel |
| 134 | - Google Analytics |
| 135 | - Other / None |
| 136 | |
| 137 | [Based on choice, provide tailored suggestions] |
| 138 | ``` |
| 139 | |
| 140 | ### Step 5: Business Outcomes |
| 141 | **Ask:** "How does this feature ladder up to company KPIs?" |
| 142 | |
| 143 | **Prompt based on company context:** |
| 144 | ``` |
| 145 | [If company detected]: |
| 146 | At [Company], typical product KPIs include [research-based suggestions]. |
| 147 | Which does this feature impact? |
| 148 | |
| 149 | [If no company]: |
| 150 | What business metric improves if this feature succeeds? |
| 151 | - User acquisition (viral, referral) |
| 152 | - Activation (first value, aha moment) |
| 153 | - Engagement (DAU/MAU, session length) |
| 154 | - Retention (churn reduction) |
| 155 | - Revenue (expansion, upsell) |
| 156 | ``` |
| 157 | |
| 158 | ### Step 6: Validation Protocol |
| 159 | **Generate based on feature type:** |
| 160 | |
| 161 | For web features: |
| 162 | ``` |
| 163 | How agents prove it works: |
| 164 | |
| 165 | 1. Functional Tests |
| 166 | - [Generate based on observable behaviors] |
| 167 | |
| 168 | 2. Performance Tests |
| 169 | - [Generate based on time constraints] |
| 170 | |
| 171 | 3. Error Handling Tests |
| 172 | - [Generate based on failure scenarios] |
| 173 | |
| 174 | 4. Analytics Validation |
| 175 | - Events fire correctly |
| 176 | - Data captured accurately |
| 177 | |
| 178 | Success Rate: [X] of [Y] checks pass = green light |
| 179 | Target: 85%+ success across 20 test runs |
| 180 | ``` |
| 181 | |
| 182 | ### Step 7: Architecture Constraints |
| 183 | **Ask:** "Are there any non-negotiable technical decisions? (frameworks, tools, platforms)" |
| 184 | |
| 185 | **Prompt:** |
| 186 | ``` |
| 187 | Examples: |
| 188 | - "Must use existing component library (don't build custom UI)" |
| 189 | - "Runs as Electron app (not browser extension)" |
| 190 | - "Uses PostHog for analytics (not Amplitude)" |
| 191 | |
| 192 | This helps agents avoid exploring paths you've already ruled out. |
| 193 | ``` |
| 194 | |
| 195 | ### Step 8: Technical Blueprint |
| 196 | |
| 197 | **Purpose:** Give the implementing agent enough architectural detail to start building without a research phase. |
| 198 | |
| 199 | **Ask:** "Let me map |
People who liked this also viewed
6 picksType
- find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 2026
- grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026
- vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026
- improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026
- setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026
- triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026
