maestro-verify

# Maestro Verify Use this when a task needs verification, proof failed, or a feature gate asks for QA evidence. Activate: `maestro hook record --event skill_activation --skill maestro-verify` ## Do 1. Start with `maestro status` or `maestro task next`. 2. Inspect the work: `maestro task show <id>` and the locked acceptance checks, or `maestro feature show <id>` for feature QA blockers. 3. Run the smallest checks that can falsify the claim from the repo root. Record exact command/manual probe and outcome. 4. Complete with matching claim and proof when the task is not yet completed: ```sh maestro task complete <id> \ --summary "<what changed>" \ --claim "<claim>" \ --proof "<observed evidence>" ``` 5. If verification fails, run `maestro query proof <id>`, repair the claim or evidence, then `maestro task verify <id>`. ## Repair - Missing claim: complete or update the task with the exact observable claim that was proven. - Missing proof: add `--proof` through `task complete`, or record manual proof: `maestro event create --task-id <id> --claim "<claim>"`. - Stale proof: rerun the falsifying checks and verify again. - Failed command: fix the work or acceptance command; do not mark verified by narration. ## Adversarial Fan-out Use when a task failed verify twice, risk is high, or many tasks support a feature ship. 1. Rubric is `acceptance.yaml` plus completion claims. Do not invent softer checks. 2. Spawn one fresh verifier pe...