Files
temporal/.claude/skills/review/SKILL.md
Stephan Behnke 98859f32f8 Review agent guidelines (#8973)
## What changed?

Added prompt ("skill") for AI review agents.

Note that the source of truth file is in `.github` since it's the only
agent that does not support file references.

## Why?

There are common review issues that cannot be linted easily that we want
to enforce without human intervention.

## How did you test it?

Added a violation and ran review agents:

### Cursor 

<img width="819" height="339" alt="image"
src="https://github.com/user-attachments/assets/655aee7f-1b0c-4e7f-8ef5-60ea543d21cd"
/>

### Copilot

<img width="827" height="354" alt="image"
src="https://github.com/user-attachments/assets/cac3a819-d115-49ff-9b0c-1a51db162720"
/>

### Claude Code via `/review`

<img width="578" height="352" alt="Screenshot 2026-01-08 at 11 21 15 AM"
src="https://github.com/user-attachments/assets/057b911d-c4e4-4eda-8a99-c66480de7bb3"
/>
2026-02-18 02:16:42 +00:00

563 B

name, description, allowed-tools
name description allowed-tools
review Review code for common issues. Use when reviewing PRs, suggesting code changes, or when user asks to review code. Read, Grep, Glob

Code Review Skill

Apply these review patterns when reviewing code or suggesting changes.

@.github/copilot-instructions.md

Review Output Format

When reviewing, organize feedback by priority:

  1. Must fix: Issues that will cause bugs or break conventions
  2. Should fix: Redundant code, naming issues, test correctness
  3. Nit: Style preferences, minor improvements