What problem does it solve?
uverify helps you validate that a previously executed change is truly correct by attacking it with adversarial checks, so you can demonstrate breakage or confidently advance to review.
Core Features & Use Cases
- Attack-checklist verification: Builds a hypothesis-driven set of checks covering happy-path, negative, invariants/assumptions, and interface mismatches.
- Evidence-first verification loop: Re-runs each attack in the current message to prevent “it should work” assumptions, capturing real outputs as evidence.
- Task-file audit trail: Writes a concise Verify summary into the task file’s ## Verify section, and routes to up:ureview or up:uexecute based on outcomes.
Quick Start
Ask the agent to run uverify against the current task’s Execute output and update docs/tasks/<slug>.md under the Verify section with the attack results and next action.