---

name: skills description: Routes audit requests to the right /nitpicker command. Use when the user wants to run one of the hostile audit commands in this repo, or asks what audit commands are available.

Skills Launcher

Everything public in this repository is one skill — nitpicker — invoked as /nitpicker <command> [extra instructions]. The authoritative command table lives in skills/nitpicker/SKILL.md (## Commands); do not duplicate it here. /nitpicker help prints it.

Routing Guide

If the user says… → run:

• "review the whole repo / audit everything / pre-release check" → /nitpicker (default audit)
• "review this code / find bugs / tear this apart" → /nitpicker review
• "review this PR / give me a PR comment" → /nitpicker pr
• "implement cr comments / address pr feedback" → /nitpicker cr
• "security audit / find vulnerabilities / check for secrets" → /nitpicker security
• "audit the tests / do the tests actually test anything" → /nitpicker tests
• "audit the types / check type safety / find the anys / are the type ignores real" → /nitpicker types
• "check the docs / find stale documentation" → /nitpicker docs
• "audit CONTRIBUTING / is our contributor guide accurate / we have no CONTRIBUTING.md" → /nitpicker contributing
• "what architecture is this / detect patterns" → /nitpicker arch-profile
• "audit the architecture / find violations" → /nitpicker arch
• "perf audit / why is this slow / will this scale" → /nitpicker perf
• "audit dependencies / prune deps" → /nitpicker deps
• "audit the licenses / license compatibility / are we GPL-contaminated / check attribution" → /nitpicker license
• "find silent failures / what errors are we swallowing" → /nitpicker errors
• "audit the CI / GitHub Actions security" → /nitpicker ci
• "audit the commits / verify conventional commits" → /nitpicker commits
• "audit the migrations / is this migration safe" → /nitpicker migrations
• "audit observability / can we debug this at 3am" → /nitpicker observability
• "does the spec match the code / is this change breaking" → /nitpicker contract
• "a11y audit / check WCAG / keyboard accessible" → /nitpicker a11y
• "privacy audit / PII audit / GDPR check" → /nitpicker privacy
• "config audit / check env vars / config drift" → /nitpicker config
• "audit the infra / Dockerfile / Terraform / k8s security / IaC misconfig" → /nitpicker iac
• "audit prompt safety / check for prompt injection / is this agent safe / LLM integration" → /nitpicker prompt-safety
• "find leaks / unclosed connections / fd leak" → /nitpicker leaks
• "i18n audit / find hardcoded strings" → /nitpicker i18n
• "find race conditions / is this thread-safe" → /nitpicker concurrency
• "close loopholes / can our rules be bypassed" → /nitpicker agent-loopholes
• "enforce hooks / harden hook coverage" → /nitpicker agent-hooks
• "audit .claude/rules / rules placement" → /nitpicker agent-rules
• "be lazy / YAGNI / find bloat / over-engineering" → /nitpicker complexity
• "find unwired code / is everything hooked up / incomplete implementations" → /nitpicker unwired
• "plan this / how should we build X / design the implementation" → /nitpicker plan
• "execute the plan / implement the approved plan / build what we planned" → /nitpicker execute-plan
• "teach me X / help me learn X / explain X so it sticks / next lesson" → /nitpicker teach
• "can we ship / release gate" → /nitpicker release-gate
• "baseline the findings / accept existing debt / only fail on new findings" → /nitpicker baseline
• "what audit commands are there / list the commands" → /nitpicker help

Rules

• Select exactly one command per request. Never chain commands yourself — the nitpicker router owns any command-to-command hand-off.
• If the request matches multiple commands, pick the most comprehensive (/nitpicker default audit covers code, architecture, and docs).

If Unclear

Run make list (or uv run scripts/list-skills.py) to print the skill and its commands with descriptions, then ask the user which one fits.