Auto PR Review
Automatically review pull requests when they are opened.
Overview
Auto PR review gives you a consistent first-pass review before a human reviewer arrives. It works best when you already know what you want the agent to prioritize.
PR Opens
A new pull request or merge request enters the review queue.
Analyze Diff
Forkline inspects changed files, context, and likely risk areas.
Apply Prompt
Your configured scope and custom prompt shape the review.
Post Findings
Feedback lands as a comment or direct follow-up commit.
Info: Auto PR review requires pull request or merge request opened events.
Enabling Auto PR Review
- Navigate to Repositories
- Find your repository
- Click Settings on the repository card
- Select the Automation tab
- Find Auto PR Review
- Set the mode to Enable
- Configure scope and action
- Click Save
Configuration Options
Scope
Choose which pull requests get automatic review:
| Scope | Reviews |
|---|---|
| Forkline-authored PRs only | PRs created by @forkline-bot from automation flows |
| All PRs | Every PR in the repository |
Tip: Start with Forkline-authored PRs only if you want signal without review noise.
Action
Choose how feedback is delivered:
| Action | Description |
|---|---|
| Comment | Post a review comment with findings |
| Commit | Push suggested changes directly to the PR branch |
Custom Prompts
Define what the AI should focus on during reviews.
Security-Focused Review
Focus on security vulnerabilities:
- SQL injection risks
- XSS vulnerabilities
- Hardcoded secrets
- Authentication issues
Prioritize critical issues. Ignore style suggestions.
Performance Review
Check for performance issues:
- N+1 queries
- Memory leaks
- Unnecessary loops
- Missing caching
Focus on database queries and API endpoints.
Documentation Review
Check for documentation issues:
- Missing README sections
- Undocumented public APIs
- Unclear function comments
Ignore internal helpers and test files.
Payment Integration Example
This PR adds a new payment integration. Focus on:
1. Security: PCI compliance, no hardcoded credentials, proper encryption
2. Error handling: Network failures, timeouts, retries
3. Testing: Happy path, error cases, edge cases
Provide severity ratings (Critical/High/Medium/Low) for each finding.
Skip Conditions
Auto PR review is skipped when:
- The PR is a draft
- The PR has the
no-auto-reviewlabel
Example Review
🔍 Auto PR Review
## Summary
This PR adds user authentication. Good implementation with a few suggestions.
## Findings
### Security
- ⚠️ Password is logged in debug mode (src/auth/login.ts:45)
- ✅ Using bcrypt for password hashing
### Performance
- 💡 Consider caching user sessions (src/auth/middleware.ts:23)
### Suggestions
1. Remove debug logging of sensitive data
2. Add rate limiting to login endpoint
---
*This is an automated review. Validate all suggestions before applying.*