rules-distilllisted

# Rules Distill Scan installed skills and this repository's mistake logs, extract principles that recur across them, and distill them into rules — appending to existing rule files, revising outdated content, or creating new rule files. Applies the "deterministic collection + LLM judgment" principle: scripts collect facts exhaustively, then an LLM cross-reads the full context and produces verdicts. **Two evidence sources, two predicates, deliberately not merged.** A principle appearing in 2+ skills answers *what is cross-cutting in the catalog*; a failure-mode key with 4+ recorded occurrences answers *what keeps breaking*. Overloading one predicate onto the other would make a four-occurrence incident invisible unless it also happened to appear in two skills. **Bridges to:** `learn` — a third source, added for harness:RM-0617, answers a different question again: *what has a `learn` record already proven true by a gate*. See [Phase 1e](#1e-collect-enforced-evidence--the-third-source) below. ## When to Use - Periodic rules maintenance (monthly or after installing new skills) - After a skill-stocktake reveals patterns that should be rules - When rules feel incomplete relative to the skills being used - **Promotion:** a key in a `MISTAKES.md` has reached the threshold and `mistake-to-gate` §11 needs the rule-text half of the promotion - **The judgment hatch:** `oops` §3 classified a condition as not mechanically checkable. That route used to end nowhere; it ends here, and it