What problem does it solve? Agents often claim work is done based only on self-written tests passing, which is weak evidence. This Skill surrounds agent work with external, signal-gated verification probes and anti-greenwash rules so done/fixed/shipping claims are backed by receipts and observed evidence. ## Core Features & Use Cases - Layered Verification: Combines the project's native test suite, change-contract red-to-green checks, stack-specific probes from a catalog, anti-greenwash rules, and human gates for high-risk work. - CLI-Driven Probes: The agentit verify command plans probes, runs runnable checks with --apply, and writes a timestamped receipt to .agentit/verify/. - Evidence-Based Checklists: Checklist probes such as acceptance criteria, Postgres RLS discipline, and auth boundaries stay pending until the agent supplies concrete evidence. - Use Case: After implementing an authentication change, run agentit verify "add login endpoint" --apply, fill in the auth-boundary checklist with observed behavior, and cite the receipt path before claiming the feature is done. ## Quick Start Ask the agent to run agentit verify with a short task description and the --apply flag, then cite the receipt before claiming the task is done.
