repo-medic-lessonslisted

# repo-medic-lessons — Distilled Lessons from /evolve This companion skill ships lessons accumulated from real projects via `/evolve`. It lives outside the main skill packages so upgrades never erase them. Consult these lessons **before** running repo-medic sub-skills to avoid pitfalls we already paid for. ## When to Use - Before running `py-improve` / `config-base` / `db-tweak` / `doc-reorg` / `vue-improve` - When you see a symptom described in any lesson (cross-stage deadlock, too many clients, import drift, etc.) - When debugging a class-of-bug that recurs across sub-skills ## When NOT to Use - Single-shot fixes → use the right sub-skill directly - New unique bugs not matching any lesson → debug fresh, then run `/evolve` to add ## Lessons index (per bucket) ### py-improve - [L1 handoff-receiver-pair-check](lessons/py-improve/handoff-receiver-pair-check.md) — implicit contracts must be made explicit; any "留给 X" must verify X's claim/filter covers it (incl. enum sync case) (#8 宁缺勿伪) - [L9 silent-fallback-grade-not-uniform](lessons/py-improve/silent-fallback-grade-not-uniform.md) — critical deps must raise, not silently fall back; grade by blast radius (#8) - [L10 perf-measure-cold-and-breakdown](lessons/py-improve/perf-measure-cold-and-breakdown.md) — cold/warm split + per-stage breakdown before tuning parameters - [L11 god-fn-split-requires-characterization](lessons/py-improve/god-fn-split-requires-characterization.md) — write characterization tests to lock undocume