PR Agent

The PR Agent validates fixes through automated testing, explores alternatives using multiple AI models, and synthesizes everything into actionable recommendations.

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#1e1e2e', 'primaryTextColor': '#cdd6f4', 'primaryBorderColor': '#45475a', 'lineColor': '#6c7086', 'secondaryColor': '#313244', 'tertiaryColor': '#181825'}}}%%
flowchart LR
    subgraph validate ["🧪 VALIDATE"]
        direction TB
        V1[Read PR context]
        V2[Find or create tests]
        V3[Verify tests catch bug]
        V1 --> V2 --> V3
    end
    
    subgraph explore ["🤖 EXPLORE"]
        direction TB
        E1[Analyze proposed fix]
        E1 --> E2a[GPT-5.2]
        E1 --> E2b[Claude Opus]
        E1 --> E2c[Gemini Pro]
        E2a --> E3[Test each fix]
        E2b --> E3
        E2c --> E3
    end
    
    subgraph synthesize ["📊 SYNTHESIZE"]
        direction TB
        S1[Compare all results]
        S2[Build consensus]
        S3[Generate review]
        S1 --> S2 --> S3
    end
    
    validate --> explore --> synthesize
    
    style validate fill:#1e1e2e,stroke:#89b4fa,stroke-width:2px,color:#cdd6f4
    style explore fill:#1e1e2e,stroke:#cba6f7,stroke-width:2px,color:#cdd6f4
    style synthesize fill:#1e1e2e,stroke:#a6e3a1,stroke-width:2px,color:#cdd6f4

Loading

Every fix is tested. The agent doesn't theorize—it implements each approach, runs tests, and reports what works.

---

Quick Start

Recommended: Use Plan Mode First

For the best results, start in plan mode to create and review a detailed plan before execution:

1. Enter plan mode: Press Shift+Tab or use /plan
2. Request a review plan:

   /plan review PR #12345 - create a detailed plan for the review
   

3. Review the plan: Copilot will create a structured plan. Review the steps and make adjustments.
4. Exit plan mode: Press Shift+Tab to switch back to execution mode
5. Execute the plan:

   proceed with the plan
   

This workflow gives you control over the review approach before any actions are taken.

Direct Invocation (Alternative)

If you prefer to start immediately without planning:

# Start GitHub Copilot CLI
copilot

# Invoke the PR agent
/agent pr

# Then ask it to work on an issue or PR
please fix issue #12345
please review PR #67890

Trigger Phrases

Phrase	Description
"Fix issue #XXXXX"	Works whether or not a PR exists
"Work on issue #XXXXX"	Investigate and implement a fix
"Review PR #XXXXX"	Review an existing PR with independent analysis
"Continue working on #XXXXX"	Resume from where you left off

Phase Details

Phase 1: Pre-Flight

Typical time: ~2-5 minutes

Context gathering only - no code analysis, no fix opinions, no running tests.

• Create state file in CustomAgentLogsTmp/PRState/pr-XXXXX.md
• Read issue description and comments
• Document platforms affected
• Identify files changed (if PR exists)
• Note disagreements and edge cases from comments

Phase 2: Tests

Typical time: ~5-15 minutes (longer if creating new tests)

Ensure reproduction tests exist - tests must be verified to FAIL.

• Check if tests already exist in PR
• If not, create them using the write-ui-tests or write-xaml-tests skill
• Verify tests compile
• Verify tests FAIL (reproduce the bug)

Phase 3: Gate ⛔

Typical time: ~5-10 minutes

Mandatory checkpoint - must pass before continuing.

• For PRs: Verify tests FAIL without fix, PASS with fix
• For Issues: Confirm tests were verified to fail in Phase 2
• Uses verify-tests-fail-without-fix skill

If Gate fails, stop and fix the tests before proceeding.

Phase 4: Fix

Typical time: ~10-30 minutes (varies by complexity)

Explore independent alternatives using the try-fix skill.

Even when a PR already has a fix, Phase 4 requires running try-fix to:

1. Generate independent fix ideas (without looking at PR's solution)
2. Test alternatives empirically
3. Compare with PR's fix

The PR's fix is already validated by Gate. The try-fix skill explores if there's something better.

Phase 5: Report

Typical time: ~2-5 minutes

Deliver the result:

• For PRs: Write review with APPROVE or REQUEST CHANGES
• For Issues: Create a new PR with the fix

Skills Used

Skill	Purpose
write-ui-tests	Creates UI tests for issues following MAUI conventions
write-xaml-tests	Creates XAML unit tests for parsing/compilation issues
verify-tests-fail-without-fix	Verifies tests catch bugs (auto-detects mode)
try-fix	Proposes and tests independent fix approaches
pr-build-status	Retrieves Azure DevOps build status for PRs

State Tracking

The agent maintains state in CustomAgentLogsTmp/PRState/pr-XXXXX.md files. This allows:

• Resuming interrupted sessions
• Tracking progress across phases
• Recording fix attempts and their outcomes

When NOT to Use

Task	Use Instead
Just run tests manually	sandbox-agent
Only write tests without fixing	write-tests-agent
Quick manual validation	sandbox-agent

File Locations

File	Purpose
.github/agents/pr.md	Phases 1-3 (Pre-Flight, Tests, Gate)
.github/agents/pr/post-gate.md	Phases 4-5 (Fix, Report)
.github/skills/try-fix/SKILL.md	Fix exploration skill
.github/skills/write-ui-tests/SKILL.md	UI test creation skill
.github/skills/write-xaml-tests/SKILL.md	XAML test creation skill
.github/skills/verify-tests-fail-without-fix/SKILL.md	Test verification skill

Related Documentation

• UI Testing Guide - How to write UI tests
• XAML Testing Guide - How to write XAML unit tests
• Contributing Guidelines