self-audit

To detect Claude Code configuration drift after cross-device sync, you need an evidence-cited audit that compares local settings.json, CLAUDE.md, and skill body hashes against a canonical setup. This process validates identity, freshness, and alignment using paths, hashes, and byte-level diffs rather than guesses.

Verifying MCP server configuration and usage tiers requires checking loaded-but-unused servers and comparing them against a documented CLI to API to MCP hierarchy. A capability and hygiene review surfaces tier mismatches, slow in-session patterns, and memory hygiene issues like stale entries or missing cross-links.

An evidence-based drift audit works by validating identity and freshness, then comparing local skills and configuration files using hashes and byte-level diffs against a canonical source. It generates a dated vault output with created, type, and device metadata to ensure environments stay aligned.

You need to run a configuration verification audit on a weekly cadence, after cross-device syncs, or following a prolonged device absence. It is also triggered when explicitly requesting a drift check to prevent silent misalignment between your local setup and the canonical configuration.

Claude Code setup supports memory hygiene checks by auditing vault memory for age, stale entries, contradictions, and missing cross-links. It writes the audit results into a dated Obsidian scratchpad note, accumulating a searchable history of capability and hygiene reviews.

Checking if documented escalation rules were applied involves confirming whether the rules were triggered and followed during the session. The audit flags any violations with required next steps and writes the actionable escalation results into a dated vault output note.