This forces you to ask "should this exist?" before polishing how it exists. When reviewing PRs that add new APIs, endpoints, parameters, or config surface, it walks you through consumer archaeology: actually grep for call sites, measure what parts of the return value consumers read (not just that they exist), and kill dimensions that encode policy instead of facts. The source cites a real example where six agents produced 30+ implementation comments on PR #15695 before anyone noticed one channel was a duplicate entry point and another should be deleted outright. Treats "no consumers yet" and "forward compatible" as red flags requiring demand justification, not implementation polish. Basically applies Musk's algorithm (question requirements, delete, then optimize) to code review so you stop writing thoughtful feedback on things that shouldn't ship.
npx -y skills add cherryhq/cherry-studio --skill demand-first-review --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
prisma/skills
firebase/agent-skills
wordpress/agent-skills
Dexploarer/hyper-forge
prisma/skills