AI ops — eval gate before the next write tool ≈ half a day

Test before the next agent tool

Golden traces, block/waive/pass, retest on scope change. Eval is a release, not a workshop.

Start as sprint plan Opens the Sprint Planner with a matching template.

1. Gate before the tool

A write tool is not a demo upgrade.
- Series: Test before the next agent tool series
- Part 1: Test first story
2. Expected and abuse cases

Expected side effects and one abuse case.
- Part 2: Golden traces story
3. Cadence, then runtime

Retest on scope change; runtime stays HITL/rollback.
- Part 3: Retest story
- Follow-on: AI foundations path