---
description: "Stage-10/11 retrospective: extract durable learnings from the run's ledger and summary, and append them ONLY via sdlc_memory — the existing learnings write path."
argument-hint: "[run-id]"
owner: RStack developed by Richardson Gunde
---

# /sdlc-retro — Run Retrospective

1. Read the run: `sdlc_status`, `.rstack/runs/<id>/plan.md`, `summary.json`, `events.jsonl` (retry/blocked/gap events are the gold), and `validation.json` verdicts.
2. Extract: what failed and why (root causes, not symptoms), what the retry budget was spent on, which specialist gaps appeared (#557 `specialist_gap` events), which decisions unblocked work.
3. For each durable learning, append via `sdlc_memory` (`action: append`) — one focused learning per call. NEVER write memory files directly; the validator-approved trust model owns admission.
4. Report the learnings appended and anything NOT worth keeping (trivia, one-time outages).
