PluginBench
Skill
Pass
Audit score 90

pi-result-handling

agents365-ai/365-skills

Internal guidance for presenting Pi helper output back to the user

What is pi-result-handling?

This is internal documentation for how coding agents should handle and present results from the Pi code review helper. It ensures Pi output is preserved accurately, findings are presented clearly with proper severity ordering, and critical safeguards like asking user permission before auto-fixing are enforced.

  • Preserves Pi helper verdict, summary, findings, and next steps structure exactly as reported
  • Presents review findings ordered by severity with file paths and line numbers intact
  • Maintains evidence boundaries by keeping inferences, uncertainties, and follow-up questions distinct
  • Enforces mandatory user confirmation before applying any code fixes from reviews
  • Handles failed or incomplete Pi runs by reporting the failure rather than attempting workarounds
  • Directs users to setup flows when Pi binary or provider configuration is missing

How to install pi-result-handling

npx skills add https://github.com/agents365-ai/365-skills --skill pi-result-handling
Claude Code
Cursor
Windsurf
Cline

How to use pi-result-handling

  1. 1.Install the skill to enable proper Pi result handling in your agent
  2. 2.When Pi helper returns output, preserve its verdict and findings structure exactly
  3. 3.Present findings ordered by severity with original file paths and line numbers
  4. 4.Keep inferences and uncertainties marked as such; do not convert them to facts
  5. 5.After showing review findings, explicitly ask the user which issues to fix before making changes
  6. 6.If Pi setup is required, direct the user to /pi:setup instead of attempting workarounds

Use cases

Good for
  • Presenting code review findings from Pi to users in a structured, actionable format
  • Ensuring code fixes are never auto-applied without explicit user approval
  • Handling Pi helper failures gracefully by reporting errors instead of guessing at solutions
  • Preserving the distinction between observed facts and inferences in security or quality reviews
  • Forwarding Pi output through companion helpers while maintaining output integrity
Who it's for
  • AI coding agents (Claude Code, Cursor) that integrate Pi code review helpers
  • Developers using Pi-based code analysis and review workflows
  • Teams requiring strict controls over when and how code changes are applied

pi-result-handling FAQ

Can I auto-apply fixes from Pi reviews?

No. You must explicitly ask the user which issues they want fixed before touching any files. Auto-applying fixes is strictly forbidden, even for obvious issues.

What should I do if Pi fails or returns incomplete output?

Report the failure with the most actionable stderr lines and stop. Do not attempt to guess or provide a substitute answer on the Claude side.

How should I handle inferences or uncertain findings from Pi?

Preserve the distinction by keeping them marked as inferences or uncertainties rather than converting them to definite facts.

What if the Pi binary is missing or no provider is configured?

Direct the user to /pi:setup and do not improvise alternate flows or workarounds.

Should I reorder or summarize Pi's findings?

No. Use file paths and line numbers exactly as reported, and preserve the helper's output sections and ordering by severity.

Full instructions (SKILL.md)

Source of truth, from agents365-ai/365-skills.


name: pi-result-handling description: Internal guidance for presenting Pi helper output back to the user user-invocable: false

Pi Result Handling

When the helper returns Pi output:

  • Preserve the helper's verdict, summary, findings, and next steps structure.
  • For review output, present findings first and keep them ordered by severity.
  • Use the file paths and line numbers exactly as the helper reports them.
  • Preserve evidence boundaries. If Pi marked something as an inference, uncertainty, or follow-up question, keep that distinction.
  • Preserve output sections when the prompt asked for them, such as observed facts, inferences, open questions, touched files, or next steps.
  • If there are no findings, say that explicitly and keep the residual-risk note brief.
  • If Pi made edits, say so explicitly and list the touched files when the helper provides them.
  • For pi:pi-companion-forwarder, do not turn a failed or incomplete Pi run into a Claude-side implementation attempt. Report the failure and stop.
  • For pi:pi-companion-forwarder, if Pi was never successfully invoked, do not generate a substitute answer at all.
  • CRITICAL: After presenting review findings, STOP. Do not make any code changes. Do not fix any issues. You MUST explicitly ask the user which issues, if any, they want fixed before touching a single file. Auto-applying fixes from a review is strictly forbidden, even if the fix is obvious.
  • If the helper reports malformed output or a failed Pi run, include the most actionable stderr lines and stop there instead of guessing.
  • If the helper reports that setup is required (pi binary missing, no provider configured), direct the user to /pi:setup and do not improvise alternate flows.