claude-md-audit

Audit and optimize an existing CLAUDE.md — cut the noise, keep the signal.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "claude-md-audit" with this command: npx skills add reizam/claude-md-templates/reizam-claude-md-templates-claude-md-audit

claude-md-audit

Audit and optimize an existing CLAUDE.md — cut the noise, keep the signal.

Works on both global (~/.claude/CLAUDE.md ) and project (./CLAUDE.md ) files. Auto-detects which type based on content and location.

Read project context for comparison For project CLAUDE.md, also read:

  • package.json or pyproject.toml — dependencies, scripts

  • README.md — project description

  • tsconfig.json or equivalent — language config

  • .eslintrc / biome.json / ruff.toml — linting rules

  • .claude/settings.json — hooks configuration

  • Top-level directory structure

Evaluate every line against 5 criteria For each line in the file, classify it:

REMOVE — line adds no value:

  • Discoverable from code, config files, or directory structure

  • Generic best practice any senior engineer knows

  • Duplicates information from README or other docs

  • Narrative or description rather than actionable instruction

  • Already enforced by a linter, type system, or hook

KEEP — line earns its place:

  • Agent would make the wrong choice without it

  • Not inferable from any file in the repo

  • Specific enough to change agent behavior

REWRITE — good intent, bad execution:

  • Too verbose (can be said in fewer words)

  • Missing the "why" (especially for "Do NOT" rules)

  • Could be merged with another line

ADD — missing high-value context:

  • Philosophy section missing or too generic

  • Architecture decisions the agent would get wrong

  • Guardrails with no explanation of why

  • Domain context not captured anywhere

Generate the audit report

Present findings in this format:

Audit: [filename]

Current: [N] lines | Target: [30 or 40] lines

Lines to REMOVE

LineContentReason
12"Use 2-space indentation"Enforced by .prettierrc
15"React 18 with TypeScript"Discoverable from package.json

Lines to REWRITE

LineCurrentSuggestedWhy
8"Do not create barrel files""Do not create barrel files — causes circular deps"Missing reason

Lines to ADD

SectionSuggestionWhy
PhilosophyAdd 3-5 engineering valuesShapes how agent thinks

Lines to KEEP

[List lines that are good — earned praise only]

Propose the optimized version Write the complete rewritten file with all changes applied.

Show the delta

Before: [N] lines | After: [M] lines | Removed: [X] | Added: [Y] Estimated token savings: ~[Z] tokens per conversation

Ask before writing Present the report and proposed version. Only write the file if the user approves. If the project would benefit from .claude/rules/ splitting, suggest it.

Audit: ./CLAUDE.md

Current: 67 lines | Target: ~40 lines

Lines to REMOVE (23 lines)

LineContentReason
3"Built with React and TypeScript"In package.json
5-14Directory tree listingAgent runs ls
18"npm install"Standard command
19"npm run dev"Standard script
25"Use meaningful variable names"Generic best practice
26"Write unit tests for new features"Generic best practice
30"Use 2-space indentation"In .prettierrc

Lines to REWRITE (3 lines)

LineCurrentSuggestedWhy
42"Do not use any""Strict TypeScript: prefer unknown over any"More specific
45"Do not modify the database schema without asking""Do not modify /lib/db/schema.ts — migration required"Name the file

Lines to ADD

SectionSuggestionWhy
Philosophy"Schema-first: Zod defines contracts"Architecture principle missing
Do NOTAdd reason to each guardrailAgent needs "why" to generalize

Before: 67 lines | After: 38 lines | Removed: 23 | Added: 2 | Rewritten: 3 Estimated token savings: ~400 tokens per conversation

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

claude-md-global

No summary provided by upstream source.

Repository SourceNeeds Review
General

claude-md-project

No summary provided by upstream source.

Repository SourceNeeds Review
Security

compliance-evidence-assembler

把审计所需证据整理成目录、清单和缺失项,便于后续评审。;use for compliance, evidence, audit workflows;do not use for 伪造证据, 替代正式审计结论.

Archived SourceRecently Updated
Security

skillguard-hardened

Security guard for OpenClaw skills, developed and maintained by rose北港(小红帽 / 猫猫帽帽). Audits installed or incoming skills with local rules plus Zenmux AI intent review, then recommends pass, warn, block, or quarantine.

Archived SourceRecently Updated