Review (Grade A)Security-tested development skill for Claude AI. Grade A. Verify the build matched the plan. Automated checks + walkthrough with you.

Review is a security-tested development skill for Claude AI, designed to verify that the build matches the planned specifications. It is graded as Grade A, indicating a high level of reliability and effectiveness. The skill utilizes automated checks, followed by a conversational user acceptance testing (UAT) walkthrough with the user, to identify any discrepancies between the planned and built phases. The Review skill operates in multiple phases, with each phase consisting of three-layer checks against must-haves. It presents findings clearly and helps the user decide what's good enough versus what needs fixes. The skill also supports various command-line arguments, such as --auto-fix, --teams, and --model, which enable features like automatic diagnosis and creation of gap-closure plans, parallel specialist verifiers, and model overrides. To initiate the Review process, the user must provide a phase number as an argument. The skill then displays a banner indicating the start of the review phase and proceeds to execute a series of automated checks and user interactions. The skill also supports event-driven auto-verification, which allows it to automatically queue verification after executor completion. One of the key strengths of the Review skill is its ability to minimize reading subagent output, reducing the amount of tokens required for verification. However, in cases where gap details are needed for inline presentation, the skill can read full VERIFICATION.md bodies if the context window tokens are sufficient. Overall, the Review skill is designed to provide a rigorous and thorough verification process, ensuring that the built phase meets the planned specifications and is free from critical issues. Its automated checks, conversational UAT, and support for various command-line arguments make it a valuable tool for developers working with Claude AI.