# Lessons Log

A ring buffer of the last 15 actionable lessons extracted from retrospectives and incidents.
Downstream skills consume this file:
- `.cursor/skills/new-task/SKILL.md` (Step 2 Complexity Assessment)
- `.cursor/skills/plan/steps/06_work-item-epics.md` (epic sizing)
- `.cursor/skills/decompose/SKILL.md` (Step 2 task complexity)
- `.cursor/skills/autodev/SKILL.md` (Execution Loop step 0 — surface top 3 lessons)

Categories: estimation · architecture · testing · dependencies · tooling · process

---

- [2026-05-16] [tooling] Environment-mismatch test skips need a same-priority "make it green in CI" follow-up ticket logged the moment the skip is introduced; deferring it leaves 12+ tests dormant across cycles and erodes the "all skips are legitimate" claim.
  Source: _docs/06_metrics/retro_2026-05-16.md

- [2026-05-16] [process] When an autodev implement batch contains zero source-code diff (orchestration artifacts only: state file, batch report, task-archive move), the per-batch `/code-review` invocation should be waived in-batch with an explicit "PASS (waived — zero net code change)" entry instead of run; the implement skill currently mandates the call unconditionally.
  Source: _docs/06_metrics/retro_2026-05-16.md

- [2026-05-16] [estimation] 1-SP structural-cleanup tasks (no behavior change, no source diff) auto-chain through the full Phase B (Test-Spec Sync / Update Docs / Security / Perf / Deploy) the same way a 5-SP feature does; consider a "structural-only" task-type marker that the orchestrator can use to short-chain straight to Retrospective.
  Source: _docs/06_metrics/retro_2026-05-16.md