Help us make the next run of this kind of job better.

Job and result: [supply the brief, result and receipt]
What happened: [a specific success, miss or correction]
What I want to improve: [one useful outcome]

Use the supplied evidence rather than guessing my preferences. Explain one concrete lesson and propose one small change to the context, brief, example or check we use next time.

Bring back:
- The observation and supporting evidence.
- The proposed change, with exact draft wording if useful.
- Where it belongs and when it applies.
- One next-run check that would tell us whether it helped.
- A reason to reject or revise the proposal if it doesn't fit.

Keep the proposal separate from adopted instructions. Don't update persistent memory, standing instructions, shared files or an automated workflow unless I explicitly ask you to adopt that specific change. Reviewing a lesson and reusing it later is different from training model weights. One good proposal is enough for this pass.
