Use when coordinating PR review work as orchestrator. Defines delegation rules, verification, and completion criteria. Trigger with /start-pr-review [PR_NUMBER].
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
Use when coordinating PR review work as orchestrator. Defines delegation rules, verification, and completion criteria. Trigger with /start-pr-review [PR_NUMBER].
This skill defines the coordination workflow for Pull Request reviews. The approach is coordination-only: monitor PR status, delegate review tasks to specialized subagents, track completion, and report results.
Core Principle: This workflow separates coordination from work. The coordinator delegates, monitors, and reports — it does not perform direct review work.
Prerequisites
GitHub CLI (gh) installed and authenticated
Python 3.8+ for running automation scripts
AI Maestro configured for inter-agent communication
Access to spawn subagents for delegation
Output
Output Type
Format
Description
Subagent Delegation
Task spawn
Spawned subagent with PR review/fix instructions
Status Report
Text/JSON
Current PR status and action recommendations
Verification Result
JSON
Pass/fail status for all completion criteria
User Notification
Text
Human-readable summary of PR readiness
Polling Schedule
Background task
Recurring PR status checks
Instructions
Poll for PRs requiring attention - Run eia_orchestrator_pr_poll.py to get list of open PRs and their status
Identify PR author type - Determine if PR is from human contributor or AI/bot (see section 5)
Classify work needed - Review PR status and determine action type (review/changes/verification/wait)
Delegate to appropriate subagent - Spawn specialized subagent based on work type (never do the work yourself)
Monitor subagent progress - Use polling to track completion (never block waiting)
Verify completion criteria - Run eia_verify_pr_completion.py before reporting ready
Report to user - Provide status update and await merge decision (never merge without approval)
Handle failures - If verification fails, identify gaps and delegate fixes (see section 6.2)
Checklist
Copy this checklist and track your progress:
Poll for PRs requiring attention using eia_orchestrator_pr_poll.py
Identify if PR is from human or AI/bot author
Classify the work needed (review/changes/verification/wait)
Delegate to appropriate subagent (never do work yourself)
Monitor subagent progress via polling (never block)
Verify all completion criteria using eia_verify_pr_completion.py
Report status to user and await merge decision
Handle any failures by delegating fixes
Follow the decision tree below to determine the appropriate action for any PR review request.
Quick Reference: Decision Tree
PR Review Request Received
│
├─► Is PR from human or AI agent?
│ ├─► Human PR → Escalate to user for guidance
│ └─► AI/Bot PR → Direct delegation allowed
│
├─► What type of work is needed?
│ ├─► Code review → Delegate to review subagent
│ ├─► Code changes → Delegate to implementation subagent
│ ├─► CI verification → Delegate to CI monitor subagent
│ └─► Status check → Use polling script directly
│
├─► Should I block waiting?
│ └─► NEVER block. Always use background tasks or polling.
│
└─► Is PR ready to merge?
├─► Run completion verification script
├─► All criteria pass → Report to user, await merge decision
└─► Criteria fail → Identify gaps, delegate fixes
Location: scripts/eia_orchestrator_pr_poll.pyPurpose: Get all open PRs, check status, identify actions needed
When to use: On each polling interval to survey PR landscape
Output: JSON with prioritized action list
Location: scripts/eia_verify_pr_completion.pyPurpose: Verify all completion criteria for a specific PR
When to use: Before reporting PR ready, before merge
Output: JSON pass/fail with detailed reasons
This workflow must NEVER execute blocking operations. All long-running tasks must be:
Delegated to subagents
Run as background tasks
Monitored via polling
Rule 2: Never Write Code
The coordinator role is to coordinate, not to implement. Code writing is ALWAYS delegated to implementation subagents.
Rule 3: Never Merge Without User
The coordinator may verify merge readiness but NEVER executes merge without explicit user approval.
Rule 4: Always Verify Before Reporting
Before reporting any status (ready, complete, blocked), always run the verification script to confirm.
Rule 5: Human PRs Require Escalation
PRs from human contributors require different handling. Always escalate to user for guidance on communication and decisions.
Examples
Example 1: Standard PR Review Coordination
# Poll for open PRs requiring action
python scripts/eia_orchestrator_pr_poll.py --repo owner/repo
# For each PR needing review, delegate to review subagent# (orchestrator spawns subagent with appropriate prompt)# Verify completion before reporting
python scripts/eia_verify_pr_completion.py --repo owner/repo --pr 123
Example 2: Verify PR is Ready to Merge
python scripts/eia_verify_pr_completion.py --repo owner/repo --pr 123
# If complete: true, report to user for merge decision# If complete: false, identify failing_criteria and delegate fixes
Error Handling
Issue: Subagent not returning results
Cause: Subagent may have crashed or become unresponsive
Solution: Check subagent logs, re-delegate with simpler scope
Issue: PR status appears stale
Cause: GitHub API rate limiting or cache
Solution: Wait briefly, then re-poll. If persistent, check API rate limits.
