novahiz-auditlisted

# novahiz-audit End-of-session audit for Novahiz. It only checks the rules for the categories actually encountered during the session. It relies on verifiable state (execution ledger, gate journal, loaded skills, session diff) and never on agent memory. ## When to run Run at the END of a session, or when the user says "audit", "verify", "compliance check", "what we forgot". Do not run mid-implementation as a substitute for convergence. `novahiz-converge` measures the gap between request and code; this skill measures session compliance with the Novahiz rules. ## What counts as evidence | Evidence | Source | |----------|--------| | Execution ledger | `novahiz_task action="status"` and `action="signals"` | | Gate journal | rules that fired during the session (matchedRules) | | Loaded skills | skills present in the session context | | Session diff | files actually modified | Agent memory is never evidence. If a claim cannot be traced to one of the four sources above, it is marked unverifiable, not satisfied. ## Procedure 1. **Detect categories.** Collect every category flagged during the session (gate reports, primary category, secondary ones). Add `general` as a fallback when nothing else applies. 2. **Load the rule set.** Read the rules for each detected category from `catalog/rules.json` and the non-optional roadmap steps from `catalog/categories.json`. 3. **Check each rule** against the four evidence sources. Status is one of: - `satisfied`: evidence exists and ma