retrolisted

# Retro You conduct a retrospective to continuously improve the workflow. ## Execution This Skill is executed by the Main Agent **itself** — not delegated to a Sub-Agent. The Main Agent experienced the workflow and can reflect on it. A Sub-Agent has no context about what the run actually felt like. **Data sources:** - **Own experience** — What went well? What was cumbersome? Where was time lost? - **plan.md** — workflow state as a memory aid for steps, results, findings - **retro_notes SQL table** — on-the-fly collected improvement notes from the running workflow. Query: `SELECT * FROM retro_notes ORDER BY id` ## Approach ### 1. Recap the workflow run - **plan.md as a data source**: Read plan.md to reconstruct the actual run — which Skills ran, which were skipped, and what results they had. Go through the completed steps: - `/analyze` → `/discuss` → `/research` → `/conceptualize` → `/design` → implementation/`/brief` - `/simplify` → `/test-review` → `/review` → `/security-review` → `/doc-review` ### 2. Evaluate each step For each Skill that ran: - What worked well? - What was cumbersome or missing? - How long did the step take (subjectively: fast/appropriate/too long)? - Was the output helpful for the next step? ### 3. Overall evaluation - Was the original problem solved? - Was the workflow run efficient? - Were there steps that could have been skipped? - Were there missing steps? ### 4. Improvement proposals Formulate concrete, actionable suggestions: - Changes