retrolisted

# /retro — Sprint Retrospective (Reflect + Act) **The retro fixes problems immediately — it does not just document them.** The live per-task mini-retros at `docs/spec/retros/sprint-S<N>-tasks.md` are the PRIMARY input. The full retro file aggregates + classifies + acts on them. The backlog audit at Step 8 is a **HARD GATE** — mismatch ≠ 0 blocks the retro close. ## Token budget (MANDATORY) - One `Read` of the live mini-retro file with `limit: 400` is the budget for the primary input. - One `Read` of the sprint file with `limit: 200` for the task table. - All other inputs via `Grep` (backlog audit, lesson recurrence, etc.). - Never full-Read every design doc — quote at most a handful of lines per task. ## Steps 1. **Read the live mini-retro file** — `docs/spec/retros/sprint-S<N>-tasks.md`. If this file is empty or missing, the sprint failed the per-task retro discipline; flag it as a process loss in the full retro. 1b. **Read `docs/spec/FOLLOWUPS.md`** — open rows are candidates for "consumed this sprint?" reconciliation in Step 8b. Note the highest existing `F####` ID so you can pick the next free one if you append. 2. **Read the sprint file** + `git log` since the sprint open date for each touched repo. 3. **Cross-cutting drift detection** — for whatever invariants apply to {{PROJECT_NAME}} (contract drift, schema drift, multi-tenancy bypass, observability gaps), grep + flag any drift. Each drift = a finding. 4. **Ask the user** (or "auto" to generate from data) for to