This is a release gate system that runs three layers of AI review before you publish to npm. Up to 10 "ultrabrain" agents review individual change groups, then 5 agents do a holistic pass on security and goal compliance, then a final oracle agent checks version bumps and breaking changes. It's overkill for small projects, but if you're shipping critical packages where a bad release costs you actual reputation or money, having 16 agents scrutinize your diff before publish might be worth the compute cost. You start by running get-unpublished-changes to establish what's actually changed since your last release.
npx -y skills add code-yeongyu/oh-my-openagent --skill pre-publish-review --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
cursor/plugins
github/awesome-copilot
alirezarezvani/claude-skills
microsoft/win-dev-skills