Mandates empirical evidence and command execution before any code task is marked as complete or fixed.
Verification Guard is a specialized Claude Code skill designed to eliminate false completion claims and hallucinated success reports. It enforces a strict 'evidence before assertions' workflow, mandating that the AI runs fresh verification commands—such as unit tests, linters, or build scripts—and analyzes the actual output before declaring a task finished. This prevents broken builds, regression errors, and shipping incomplete features by treating verification as a non-negotiable gate function, ensuring that every claim of success is backed by verifiable logs and exit codes.
主要功能
01Mandatory fresh command execution before any status updates
02Elimination of vague or speculative language regarding code health
03Systematic requirement checklist verification to prevent feature gaps
04Independent validation of agent-delegated work via VCS diff analysis
05Automated Red-Green regression testing cycles to confirm bug fixes
060 GitHub stars
使用场景
01Validating build success before pushing to production or creating Pull Requests
02Confirming bug fixes with actual passing test results rather than assumptions
03Ensuring linter compliance and code style adherence across entire files
