跳到正文
cccskills
免费在 GitHub 公开

vc-audit-context

Audit project context routing, shared-skill discoverability, and Claude/Codex wiring. Use when context docs or skill surfaces move, split, or drift.

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(12)

  • SKILL.md4.5 KB
  • references/audit-context.md5.3 KB
  • references/skill-routing-policy.json153 B
  • scripts/generate-skills-catalog.mjs2.2 KB
  • scripts/shared-skill-utils.mjs5.3 KB
  • scripts/validate-confusable-skills.mjs1.3 KB
  • scripts/validate-context-discovery.mjs14.6 KB
  • scripts/validate-protocol-discovery.mjs4.0 KB
  • scripts/validate-skill-cross-refs.mjs820 B
  • scripts/validate-skill-dependencies.mjs1.2 KB
  • scripts/validate-skill-keywords.mjs2.0 KB
  • scripts/validate-skill-routing.mjs3.4 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Audit Context

Output style: Follow process/development-protocols/communication-standards.md — answer-first, plain language, no unexplained jargon, TL;DR on long responses.

Use this skill to verify that the project's durable context layer is discoverable and organized.

Optional input: a context group, agent, skill, or folder scope to prioritize during the audit.

Workflow

  1. Run find process/context/ -type f | sort to get the full file listing before routing. This ensures no context file is silently skipped when the router is incomplete or drifted.
  2. Read process/context/all-context.md for the context routing protocol.
  3. Read references/audit-context.md for the full audit process.
  4. Run the context discovery validator:
    node .claude/skills/vc-audit-context/scripts/validate-context-discovery.mjs
    

3a. Run the protocol discovery frontmatter validator (enforces discovery frontmatter on every process/development-protocols/**/*.md, recursive incl. vc-system-behavior/; note.md is the only intentional exclusion):

node .claude/skills/vc-audit-context/scripts/validate-protocol-discovery.mjs
  1. Run the shared skill routing coverage validator:
    node .claude/skills/vc-audit-context/scripts/validate-skill-routing.mjs
    
  2. Run the skill cross-reference validator:
    node .claude/skills/vc-audit-context/scripts/validate-skill-cross-refs.mjs
    
  3. Run the skill dependency/confusable analysis:
    node .claude/skills/vc-audit-context/scripts/validate-skill-dependencies.mjs
    node .claude/skills/vc-audit-context/scripts/validate-confusable-skills.mjs
    
  4. Regenerate or check the machine-readable skill catalog:
    node .claude/skills/vc-audit-context/scripts/generate-skills-catalog.mjs --write
    node .claude/skills/vc-audit-context/scripts/generate-skills-catalog.mjs --check
    
  5. Validate that every SKILL.md carries trigger_keywords + a valid layer (contract|helper) and that the catalog is in sync:
    node .claude/skills/vc-audit-context/scripts/validate-skill-keywords.mjs
    
  6. If any script reports failures, inspect the referenced files and patch the smallest relevant surface.
  7. Re-run the failed validators until they pass.

For agent/skill harness validation (agent parity, skill frontmatter, README.md sync, protocol wiring), use the audit-vc skill.

Context Bootstrap (when process/context/ doesn't exist or needs full init)

Use when initializing a new project's context layer from scratch:

  1. Run vc-scout in parallel across major source directories (skip .git, node_modules, .claude, caches) to gather codebase summaries.
  2. Create process/context/all-context.md (routing table, architecture, conventions) and group all-{group}.md entrypoints for any durable domains identified.
  3. Parallel reader strategy for existing context files — before updating, spawn subagents proportional to file count: 1-3 files read directly; 4-6 files use 2-3 reader agents; 7+ files use 4-5 reader agents (max 5), distributing by LOC.
  4. After generating or updating context files, run find process/context -name '*.md' -print0 | xargs -0 wc -l | sort -rn — files over 800 LOC should be split into a context group or the user asked.
  5. Finish by running the discovery validator (step 3 above) before declaring done.

Rules

  • Treat .claude/skills/ as canonical; .agents/skills/ is the Codex discovery symlink.
  • Treat .claude/skills/vc-audit-context/references/skill-routing-policy.json as the explicit allowlist for intentionally non-routed shared skills.
  • Do not move large context files without updating process/context/all-context.md.
  • Do not delete compatibility wrappers unless no current reference points to them.
  • Keep context groups durable-domain based, not one group per temporary feature.
  • When updating agents, mirror Claude markdown and Codex TOML surfaces together.
  • Treat validator warnings as audit findings unless the user asks for a strict cleanup.
  • Prefer validator-backed routing truth over adding more soft prose.
  • Treat process/context/generated-skills-catalog.json as the machine-readable catalog owned by audit-context.

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

AI-optimized browser automation CLI with context-efficient snapshots. Use for long autonomous sessions, self-verifying workflows, video recording, and cloud browser testing (Browserbase).

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

Evaluate 4 execution strategies (sequential, parallel-subagents, workflow, agent-team) for a phase or fan-out task. Outputs 7-signal score table, agent count math, cost guards, and strategy recommendation.

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

Audit active project plan files for staleness, completion, and routing truth. Use when cleaning up plans, reconciling active work, or archiving completed artifacts.

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

Audit agent harness health: Claude/Codex agent parity, skill registry consistency, README.md sync, and protocol file wiring. Use when agents, skills, README.md, or development-protocol files move, split, or drift.

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

Emit and validate the provisional goal block for Autopilot Mode. Owns the 9-field format and resume detection from a pasted goal block.

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

Loop: find gaps → fix → repeat until agents find no gaps or a metric goal is hit. Shared loop primitive for PVL, EVL, and standalone quality runs.

中文简介正在准备中。当前显示原文说明。

withkynam/vibecode-pro-max-kit11532026年6月21日 更新

查看 withkynam 的全部技能

报告此技能的问题