Memory Hygiene Audit
Home

CLAUDE.md audit

CLAUDE.md Audit for Clean Agent Instructions

A CLAUDE.md audit focuses on the instruction file that shapes agent behavior inside a repo. The useful outcome is not a style review. It is a safer, shorter, more current operating brief that tells the agent what to do, what to avoid, and when a rule should expire.

View pricing plans

When this matters

  • The file has grown into a transcript of old fixes instead of a maintainable agent guide.
  • Multiple contributors added conflicting instructions for tests, deployment, or approvals.
  • A platform team wants a clean handoff file before broad agent rollout.

How to run the audit

  1. Parse sections for commands, preferences, policies, paths, dates, and external links.
  2. Detect contradictions, stale instructions, and unscoped permissions.
  3. Mark rules as durable, temporary, personal, project, or unsafe.
  4. Suggest a compact replacement structure for the next run.
  5. Create an evidence receipt showing what changed and why.

Common risks

  • Agents can treat old local paths or temporary flags as current instructions.
  • Unbounded language can encourage risky file or network actions.
  • Private details may be retained longer than intended.

How Memory Hygiene Audit connects this to checkout

Memory Hygiene Audit gives CLAUDE.md owners a before-and-after context diff, risk flags, and a checkout path for exportable remediation plans.

Teams can preview the score, then use the Team Hygiene annual checkout to generate the full cleanup diff, reviewer notes, and agent-readable JSON policy.