--- name: dogfood-skills description: | Enforces dogfooding for the skills CLI project. Use when working on packages/skills-cli or packages/skills-library. Reminds Claude to run `skills scan` and install recommended skills after completing features. Triggers on: feature completion, session end, or when reminded. category: development user-invocable: true --- # Dogfood Skills CLI You are working on the skills CLI project. You MUST use the tools you build. ## Core Principle This project builds tools for discovering and installing Claude skills. If you don't use these tools yourself, you: 1. Miss bugs and UX issues 2. Don't understand user pain points 3. Fail to validate the value proposition 4. Undermine the project's credibility ## Mandatory Actions ### After Completing Any Feature/Bugfix 1. **Run scan**: `./packages/cli/bin/skills.js scan` 2. **Review recommendations**: Look at HIGH CONFIDENCE matches 3. **Install relevant skills**: `./packages/cli/bin/skills.js scan --all` 4. **If skipping**: Document reason in your response ### Rationalizations (Do Not Skip) | Rationalization | Why It's Wrong | Required Action | |-----------------|----------------|-----------------| | "I'll install later" | You said that last time. And the time before. | Install NOW | | "I forgot" | Hook exists to remind you. Pay attention. | Run scan immediately | | "Not relevant to current work" | Scan output shows relevance scores | Justify with evidence from scan output | | "Too many skills" | Deduplication exists | Use `--show-alternatives` to choose best | | "Just a small change" | Small changes compound | Scan takes 10 seconds | | "Tests are passing" | Tests don't verify dogfooding | Run scan anyway | ## Blocking Condition If you complete a feature without running `skills scan`: **BLOCKED: DOGFOODING REQUIRED** You cannot proceed to the next task until you: 1. Run `./packages/cli/bin/skills.js scan` 2. Acknowledge recommendations (list them) 3. Install OR justify skipping each HIGH confidence recommendation ## When to Run Scan - After implementing a new feature - After fixing a bug - After refactoring code - At the end of a work session - When the user mentions skills or recommendations - When this skill is invoked via `/dogfood-skills` ## Quality Check At end of each session, ask yourself: - Did I run `skills scan`? - Did I review the recommendations? - Did I install recommended skills? - Did I use installed skills during work? If any answer is "no", run the scan now before ending. ## Commands Reference ```bash # Scan for recommended skills ./packages/cli/bin/skills.js scan # Scan and show all matches (not just top per category) ./packages/cli/bin/skills.js scan --show-alternatives # Install all high-confidence recommendations ./packages/cli/bin/skills.js scan --all # List currently installed skills ./packages/cli/bin/skills.js list # Get help ./packages/cli/bin/skills.js --help ``` ## Integration with Other Skills This skill works alongside: - **claudeception**: Extract learnings into new skills - **tdd**: Write tests before fixes (RED → GREEN → REFACTOR) - **unit-test-workflow**: Generate comprehensive tests After installing new skills, USE them in your workflow.