What problem does it solve? AI agents often claim they completed actions like closing GitHub issues, making commits, or creating files without actually verifying the result, which erodes trust and hides silent failures. This Skill enforces empirical verification before any completion claim. ## Core Features & Use Cases - Mandatory Verification Commands: Maps each action type to a concrete check, such as gh issue view <N> --json state for issues, git log --oneline -1 for commits, and test -f <path> for file creation. - Evidence-Based Reporting: Requires every completion report to include the actual command output as proof, replacing bare assertions like "done" or "feito". - Draft-First Integration: Combines with draft-first governance so external actions follow a draft, approve, execute, verify, and report sequence. - Use Case: After closing GitHub issue #48, the agent runs gh issue view 48 --json state, confirms the state is CLOSED, and reports the result with the command output as proof. ## Quick Start Ask the agent to close a GitHub issue or create a file and require it to show the verification command output proving the action succeeded.