skills.rest
SebasAren
SebasArenCOMMUNITY
Explore all skills by this author →
GitHub

knowledge-audit

Audit agent-facing knowledge files for staleness, accuracy, and usefulness.

2|Updated Jun 7, 2022
Search Tags:#audit#readme#conventions#docs#agents.md#rules#knowledge-base
One-click install
npx skills add https://github.com/SebasAren/SebbaFlow --skill knowledge-audit-sebasaren
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: knowledge-audit
Source: https://github.com/SebasAren/SebbaFlow/tree/main/.pi/skills/knowledge-audit
Command: npx skills add https://github.com/SebasAren/SebbaFlow --skill knowledge-audit-sebasaren

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Review the staleness and usefulness of any agent-facing knowledge in the repo: rule files (.claude/rules/), AGENTS.md (at any depth), CONVENTIONS.md, README.md, and domain docs if present. Use this skill whenever the user wants to audit, clean up, or prune their knowledge base: "audit rules", "audit docs", "audit knowledge", "review rules for staleness", "clean up the rules", "are any docs outdated?", "prune old knowledge", "check if rules are still relevant", "knowledge audit", or any variation. Also trigger when the user says they're confused by a rule/doc that seems wrong — use that as the starting point for a broader audit.

Core Features & Use Cases

Audit every agent-facing knowledge file in the repo for staleness, accuracy, and usefulness. Stale knowledge is worse than no knowledge — it sends sessions in the wrong direction. Knowledge files include:

  • .claude/rules/*.md — auto-injected rules
  • AGENTS.md at any depth (root and per-package)
  • CONVENTIONS.md — code-style source of truth
  • README.md — human-facing but agents read it too
  • docs/domain/*.md if the repo uses that pattern

Quick Start

Run the knowledge-audit skill to scan and report on all agent-facing knowledge files, with optional scope modifiers for rules, docs, or specific AGENTS.md entries.

Frequently Asked Questions about knowledge-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit agent-facing knowledge files for staleness and accuracy?▼

Auditing agent-facing knowledge files for staleness involves scanning .claude/rules, AGENTS.md, CONVENTIONS.md, README.md, and docs/domain/*.md to verify accuracy and usefulness, yielding actionable cross-file clustering recommendations.

What is an agent-facing knowledge base audit and when do I need it?▼

An agent-facing knowledge base audit reviews rule files and docs like AGENTS.md and CONVENTIONS.md to identify outdated, inaccurate, or unhelpful content. You need it when stale knowledge sends agent sessions in the wrong direction or rules seem confusing.

How do I clean up and prune outdated rules and docs in my repository?▼

Cleaning up and pruning outdated rules and docs involves applying an audit workflow with discovery, per-file checks, and cross-file clustering across .claude/rules and AGENTS.md, generating actionable recommendations to remove stale knowledge.

Does the knowledge audit work with AGENTS.md files at any directory depth?▼

Yes, the knowledge audit works with AGENTS.md files at any directory depth, including root and per-package levels. It also evaluates auto-injected .claude/rules, CONVENTIONS.md, README.md, and docs/domain/*.md patterns if present.

Why should I review repository conventions and rules for staleness?▼

Reviewing repository conventions and rules for staleness is critical because stale knowledge is worse than no knowledge—it actively sends agent sessions in the wrong direction. Auditing verifies that CONVENTIONS.md and .claude/rules remain accurate and useful.

What is the best way to check if repository docs are still relevant for AI agents?▼

The best way to check if repository docs are still relevant for AI agents is to run a comprehensive knowledge audit that cross-references README.md, AGENTS.md, and domain docs for accuracy, staleness, and usefulness, yielding actionable recommendations.

Related Skills

Explore More Content & Communication Skills→
kaptinlinkaptinlin
23

agent-md-writing

Generate CLAUDE.md and AGENTS.md guidelines from repository structure and docs.

Community
Advanced
zhu637882-stackzhu637882-stack

learn

Consolidate session learnings into a searchable repository with export formats.

Community
Intermediate
ArogyaReddyArogyaReddy

python-code-style

Enforce Python code style and documentation standards with ruff and mypy.

Community
Intermediate
seanGSISGseanGSISG
1

rules-auditor

Audit Claude Code memory files for staleness, drift, and quality issues.

Community
Advanced