This is basically three protocols bundled together to prevent the "looks good to me" problem in AI-assisted development. It covers how to receive code review feedback without reflexive agreement (technical evaluation over "great point!"), when to request systematic reviews via a code-reviewer subagent (after each task in subagent workflows), and verification gates that require actual command output before claiming anything works. The verification gate thing is the most immediately useful: it forces you to run the tests and read the output before saying tests pass. If you're doing subagent-driven development or find yourself nodding along to review comments without actually checking if they're right, this gives you concrete protocols to follow.
npx skills add https://github.com/unclecatvn/agent-skills --skill code-review