reflectlisted

# Reflect (ทบทวน) **Habit**: H7 — Sharpen the Saw | **Anti-pattern**: All output, no capability improvement — repeating the same mistakes ## Why This Exists DORA research shows per-task micro-retros (3-5 questions, 5 minutes) are more effective than monthly hour-long retrospectives. The key differentiator: teams that assign action items with owners improve; teams that just "discuss" don't. Q6 below extends the DORA pattern with a skill-effectiveness signal specific to this plugin (H7 applied to the plugin itself, per Issue #92). ## Process Ask these 6 questions. Keep answers brief — this should take no more than 5 minutes. ### 1. What went well? Reinforce good practices. What should we keep doing? ### 2. What surprised me? Surface hidden complexity. What did we not expect? Sub-prompt: did a **prior diagnosis** or shipped **fix prove wrong** this session — what evidence **reversed** it? A root cause that was confident-but-wrong is a high-value surprise to capture; this signal feeds `SKILL-EFFECTIVENESS.md`. See [`independent-source-verification.md`](../../guides/independent-source-verification.md). ### 3. What would I do differently? Improve next iteration. If starting over, what would change? ### 4. What reusable pattern did I discover? Extract for the team. Is there a script, template, or approach worth sharing? Sub-prompt: did the AI workflow itself reveal a reusable cost/health pattern — repeated loops, retries without new evidence, context compaction that ne