/reflect

description: Reflect on recent work — log lessons to the journal and propose distilled rule updates for review

Run Phase 7 (Reflect & Evolve) from `.agents/self-evolution.md`.

1. Review what changed this session: `git --no-pager status` and `git --no-pager diff` on the vault, the conversation, and especially any **corrections** the user made (corrections are the highest-value signal). 2. **Append** a dated entry to `.agents/learned/journal.md` (append-only — never rewrite past entries). Create `.agents/learned/` first if it does not exist. 3. Identify durable lessons (ones that would change a *future* run). For each, pick the right home:

• a preference/override → edit `.agents/learned/conventions.md` (consolidate;

keep it a tight rule set, not a changelog);

• a classification judgment → add/update a row in `.agents/learned/examples.md`;
• a whole new input shape → write a playbook to

`.agents/learned/skills/<kebab-name>.md` with a `trigger:` line and steps. 4. The journal entry can be committed freely. For `conventions.md` / `examples.md` / `skills/` changes, show the proposed diff, summarize each change in one line with the motivating journal entry, and let me approve before committing.