<!--
  miss-log.md — one line per thing your agent got wrong.
  Started in Session 02, mined in Sessions 03 and 04.
  Do NOT patch rules mid-week. Collect first, patch later — you're
  building a dataset about your own agent, not firefighting.

  Gap type (fill in after Session 03 teaches the three kinds):
    RULE     - no rule covered the case        -> write the rule
    MEMORY   - the agent couldn't know a fact   -> add a memory file
    JUDGMENT - rules + facts existed, wrong call -> tighten its authority
-->

# Miss Log

| Date | What happened | Bin it chose | Bin that was right | Gap type |
|------|---------------|--------------|--------------------|----------|
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |
|      |               |              |                    |          |

## Weekly tally (bring to the next session)

- Total misses this week: ______
- Rule gaps: ______   Memory gaps: ______   Judgment gaps: ______
- Misses last week (for the delta): ______

## Notes
<!-- Patterns you notice. A name that keeps getting mis-binned is a memory
     fact waiting to be written. A category that's always wrong is a rule
     waiting to be tightened. -->