post-audit-memory-capturelisted

## When to use After completing any multi-cycle adversarial analysis, security audit, or hardening pass on a component, skill, or enforcement mechanism. Trigger when: - A red-team/blue-team cycle has closed findings on a component - A skill or gate mechanism has been formally reviewed and certified - You want a future session to audit proposed changes against a known-clean baseline ## The pattern Two markdown sections per audited component, appended to the memweave corpus. No more, no less. Both go in one file so they index together and surface as a pair in `mw_search.py`: Run via the Bash tool (the tool result confirms the write). The heredoc terminator is a unique token — keep it as-is and do not let baseline content contain that exact line: ```bash mkdir -p ~/.uncle-j-memory/memory cat >> ~/.uncle-j-memory/memory/audit-baselines.md <<'PM_AUDIT_BASELINE_EOF' ## [AUDIT BASELINE] <Component> — <YYYY-MM-DD> ### Design invariants + audit baseline Invariants (numbered — the non-negotiables): 1. ... 2. ... Audit certification: <N>-cycle adversarial analysis completed <date>. Final cycle result: <only MEDIUMs/LOWs | no new CRITs or HIGHs>. Confidence baseline: any single-session edit touching invariants requires its own adversarial pass. ### Known closed attack vectors - **<Attack name>** — <one-line vector description>; closed by <one-line fix description> PM_AUDIT_BASELINE_EOF ``` ### Section 1 — Design invariants + audit baseline Properties that **must hold** in any f