retro

Test-driven development, red then green, one vertical slice at a time. Use when building a feature, a rule, a config option or a bug fix test-first, when the user says "TDD", "red-green" or "write the test first", or when a subagent is handed a ticket to build.

Prove a change before claiming it is done - judge lint, typecheck, tests and git push by their exit codes rather than by reading output, and show that a bug fix's regression test fails without the fix. Use before saying "tests pass", "lint is clean", "pushed" or "fixed in <sha>"; before opening or updating a pull request; whenever a command is piped into grep, tail or head; and whenever a bug fix or false-positive fix ships with a new test. For whether a measured number (a score, a timing, an eval result) can be credited to a change, use `explain-the-number`.

Investigate a question against primary sources and capture the findings as one cited Markdown note. Use when the user wants a topic researched, a vendor's docs or API facts gathered, a library's or tool's real behavior confirmed, or reading legwork delegated while the main session keeps working.

Runs Cloudflare's security-audit skill in full audit mode from Claude Code. It builds each agent's prompt file from that skill's text, pulls each agent's JSON reply out of its transcript, and keeps the coverage ledger, findings, agent-call budget and NEEDS-VALIDATION.md. Use it together with `security-audit` when the user asks for a full, end-to-end or report-producing security audit or pen test of a codebase. For a security question or a focused review, use `security-audit` alone.

Find skills whose descriptions fight over the same request, so Claude picks one at random or the wrong one, and fix the descriptions. Use when the user asks to audit, clean up or deduplicate their skills, when a skill fires that should not have, after installing a plugin, or before adding a new skill to a large library.

Route a multi-step task across cheaper subagents while the main session keeps the plan, the judgment calls and the verification. Use at the start of any task with more than one step, before spawning a subagent, when picking a model for a subtask, when a subagent fails, or when the user mentions cost, tokens or model choice.

Shorten an existing ticket that is too long or too technical, moving the engineering detail into a Technical notes comment so nothing is lost. Use when the user gives a ticket key or link and asks to trim, condense, simplify or clean it up, or says a product owner or analyst found it hard to read.

Writing documents for agents. Use when creating or editing a skill (its SKILL.md, description or reference files), when modifying AGENTS.md or CLAUDE.md, or when reviewing an agent-facing document that behaves inconsistently across runs. Prose that people read, such as a memo or a one-pager, follows its own rules.

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.

Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".

Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.

This skill should be used when the user wants to "write agent code", "build an agent with ADK", "add a tool", "create a callback", "define an agent", "use state management" — in a project that needs ADK (Agent Development Kit) API patterns and code examples. It provides a quick reference for agent types, tool definitions, orchestration patterns, callbacks, state management, the graph Workflow API, and reference recipes to study. Do NOT use for scaffolding (use google-agents-cli-scaffold) or deployment (use google-agents-cli-deploy).

Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.

Generate AI music on RunComfy via the `runcomfy` CLI — a smart router across the music-model catalog. Routes to ElevenLabs AI Music Generation (premium 44.1 kHz stereo vocal tracks, 5 s–5 min, $0.0083/s) and ACE Step / ACE Step 1.5 (StepFun-AI open-weights, tag-driven composition, multilingual lyrics, $0.0002–0.0003/s, ~27× cheaper), plus ACE Step audio-inpaint (regenerate a time range inside an existing track) and ACE Step audio-outpaint (extend a track before or after). Picks the right model for the user's actual intent — premium vocal hook, cheap background music library, multilingual pop song, repair a bad chorus, lengthen a 30 s draft into a 2 min cut — and ships each model's documented prompting patterns plus the minimal `runcomfy run` invoke. Triggers on "generate music", "make a song", "AI music", "background music", "instrumental track", "soundtrack", "jingle", "theme music", "royalty-free music", "compose", "music with lyrics", "extend music", "fix this song", "inpaint music", or any explicit ask to generate or edit music.

创建 lark-cli 的自定义 Skill。当用户需要把飞书 API 操作封装成可复用的 Skill（包装原子 API 或编排多步流程）时使用。