Pi Result Handling

Internal formatting rules for how Claude should present output from the Pi code review helper. The critical bit: after showing review findings, Claude must stop and explicitly ask which issues to fix rather than auto-applying changes. It preserves the helper's verdict structure, keeps findings ordered by severity, maintains the distinction between facts and inferences, and reports failures directly instead of improvising workarounds. If Pi says setup is needed, it redirects to the setup command. This is essentially a guardrail to prevent Claude from silently "fixing" things after a review when the user might want to review findings first or handle them selectively.

Install to Claude Code

npx -y skills add agents365-ai/365-skills --skill pi-result-handling --agent claude-code