跳到正文
cccskills
免费在 GitHub 公开

code-hygiene-cleanup

Execute a bounded ClawJS/Clawix code hygiene cleanup campaign from an audit report, one category and repo at a time.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md2.2 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

code-hygiene-cleanup

Use this skill when the user explicitly asks to clean code hygiene findings or continue the code hygiene goal.

Procedure

  1. Start from a current code-hygiene-audit report and the 33-decision checklist.
  2. Choose exactly one repo and one category for the batch: dependencies, orphan files, private exports/types, tests/fixtures, assets/SVG, Swift report-only review, TODOs, or duplicates.
  3. Re-run or inspect scanner output after entrypoint calibration. For Knip, separate clear private debt from public API/protocol/UI/icon barrels and framework route/config false positives.
  4. Preserve public API, package entrypoints, CLI/router/registry/protocol surfaces, canonical docs, fixtures, stable registries, and compatibility names unless an explicit removal decision exists.
  5. Make the smallest cleanup that removes proven debt. Future intent must move to issue/backlog/ADR rather than dormant code.
  6. When retaining a finding, add or update a baseline entry with reason, owner area, reference, and expiry; never hide a real bug behind EXTERNAL_PENDING.
  7. For large generated/baseline refreshes, record generated provenance with generator, command, source, upstreamHash, regenerationMode, deltaSummary, debtImpact, and conditional debtImpactReason. The delta must explain what changed and the debt impact must say whether cleanup debt decreases, stays neutral, or increases.
  8. Run the relevant checks for the batch before moving on. Record validation and baseline removals/additions in the ledger.

Constraints

  • Autofix is suggestion-only by default; do not run destructive fixes or remove files without explicit task scope.
  • Keep enum members, Swift dynamic-use findings, and duplicates report-only until reviewed.
  • Do not use @generated, a baseline filename, or generatedPaths alone to hide a large generated/baseline change.
  • Do not batch unrelated cleanup categories together.
  • Do not modify unrelated dirty work while cleaning; record it separately if it affects validation.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Govern Clawix accessibility changes across screen reader behavior, keyboard navigation, focus order, contrast, reduced motion, text scaling, timed interactions, and agent-generated UI.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

Review or update stable, canonical, any-human, PMF, and adoption claims using adoption/canonicity governance packets.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

Turn a new or changed ADR into implemented behavior, routing docs, tests, and guardrails rather than leaving it as standalone prose.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

Repair confirmed architecture drift in small batches while preserving public canon and avoiding broad opportunistic refactors.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

Given an existing Style, generate a human-readable brand guidelines document covering colors, typography, voice, do/don't, imagery rules, and usage examples. Use when the user asks for a brand book, brand guide, or design system document.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

Add or improve canonical data catalog collections, fields, aliases, relations, evidence tags, docs, and tests.

中文简介正在准备中。当前显示原文说明。

clawic/Clawix82026年6月6日 更新

查看 clawic 的全部技能

报告此技能的问题