retrolisted

# Retrospective Close the agile loop: review what happened, capture learnings, plan the next cycle. See `docs/agents/workflow.md` for Scrum vs Kanban context. ## Process Ask **one topic at a time** (grill-me-lite). Provide a recommendation based on git/issue history before each question. 1. **What shipped** - Explore: merged PRs, closed issues, tags since last retro (or last 1–4 weeks) - If `$AI_KIT_ROOT/bin/usage-stats.sh` is available and the log has events, also run `usage-stats.sh --since=7d` (or `30d`) and surface: top skills used, skills with high abandon-%. Mention this BEFORE asking open questions — it grounds the conversation in observed behaviour. - **Finding:** list increments delivered + observed skill usage - **Question:** anything missing from this list? Any skill that felt off? 2. **What blocked us** - Explore: open issues, reopened bugs, long-running PRs - **Finding:** patterns (unclear requirements, missing tests, infra, dependencies) - **Question:** top blocker to address next cycle? 3. **Feedback** - **Question:** user/stakeholder feedback (you provide — agent does not invent user sentiment) - Capture verbatim or summarised 4. **Learnings** - Should anything update `CONTEXT.md` or a new ADR? - If usage-stats showed a high abandon-% on a skill: ask why the user backed out, and whether the skill itself needs adjusting (e.g. open an issue with template `skill-suggestion`). - Only write docs if the user agrees 5. **N