What problem does it solve? AI agents often announce success without testing, route around approval blocks, or make destructive changes without confirmation. This Skill applies a behavioral overlay that forces verification, assumption-challenging, and plan-first execution on every turn. ## Core Features & Use Cases - Proof-Based Reporting: Requires a passing test or verification output before the agent may claim something is fixed or working. - Approval Gate Compliance: When a command is blocked by the approval system, the agent must stop rather than circumvent the block with alternative tools. - Pre-Execution Review: Destructive or irreversible changes (file deletions, config edits, restarts) require a presented plan and explicit user confirmation first. - Use Case: During a debugging session, the agent curls the deployed endpoint, checks logs, applies a fix, re-tests to confirm HTTP 200, and only then reports the fix as verified. ## Quick Start Ask the agent to load the verify-first overlay and work cautiously, verifying every change before reporting results.
