second-brain-audit

Run the audit on your notes folder to extract every current-sounding claim from the files your agent loads each session, then check each against the newest evidence. Contradicted claims are reported first, followed by unsupported ones.

Audit the always-loaded file such as CLAUDE.md or MEMORY.md, since a stale fact there affects every session. Then split facts into a Current State section that gets replaced on update and an append-only Log for events.

No, the audit.py script only reads and reports; it never writes to any file. Any restructuring happens separately, one page at a time, with a diff shown for your approval before anything is changed.

The script detects subjects given different monetary values in more than one file, comparing recurring rates and one-time amounts separately. Lifecycle conflicts like a page saying launching while a log records cancellation require manual review, since no regex can judge them reliably.

A zero is not a clean bill of health: the script can only compare monetary values, so notes without money are largely invisible to it. It prints a coverage warning in that case, and the manual claim-by-claim audit still applies.

Yes, daily notes are treated as a correct event log; the fix is adding one file holding current values, not restructuring the journal. For non-markdown tools, the state-versus-event idea still applies but no files are restructured.