leo-claude-mktplace/plugins/pr-review

pr-review

Multi-agent pull request review with confidence scoring and actionable feedback.

Overview

pr-review conducts comprehensive code reviews using specialized agents for security, performance, maintainability, and test coverage. Each finding includes a confidence score to reduce noise and focus on real issues.

Commands

Command	Description
/pr-review <pr#>	Full multi-agent review
/pr-summary <pr#>	Quick summary without full review
/pr-findings <pr#>	Filter findings by category/confidence
/pr-diff <pr#>	View diff with inline comment annotations
/initial-setup	Full interactive setup wizard
/project-init	Quick project setup (system already configured)
/project-sync	Sync configuration with current git remote

Review Agents

Agent	Focus
Security Reviewer	Injections, auth, data exposure, crypto
Performance Analyst	N+1 queries, complexity, memory, caching
Maintainability Auditor	Complexity, duplication, naming, coupling
Test Validator	Coverage, test quality, flaky tests

Confidence Scoring

Findings are scored 0.0 - 1.0:

Range	Label	Action
0.9 - 1.0	HIGH	Must address
0.7 - 0.89	MEDIUM	Should address
0.5 - 0.69	LOW	Consider addressing
< 0.5	(suppressed)	Not reported

Installation

Add to your project's .claude/settings.json:

{
  "plugins": ["pr-review"]
}

Requires Gitea MCP server configuration.

Configuration

Environment variables can be set in your project's .env file or shell environment.

Variable	Default	Description
PR_REVIEW_CONFIDENCE_THRESHOLD	0.7	Minimum confidence score (0.0-1.0) for reporting findings. Findings below this threshold are filtered out to reduce noise.
PR_REVIEW_AUTO_SUBMIT	false	Automatically submit review to Gitea without confirmation prompt

Example Configuration

# Project .env file

# Only show high-confidence findings (MEDIUM and HIGH)
PR_REVIEW_CONFIDENCE_THRESHOLD=0.7

# Auto-submit review to Gitea (default: false)
PR_REVIEW_AUTO_SUBMIT=false

Confidence Threshold Details

The confidence threshold filters which findings appear in review output:

Threshold	Effect
0.9	Only definite issues (HIGH confidence)
0.7	Likely issues and above (MEDIUM+HIGH) - recommended
0.5	Include possible concerns (LOW+MEDIUM+HIGH)
0.3	Include speculative findings

Lower thresholds show more findings but may include false positives. Higher thresholds reduce noise but may miss some valid concerns.

Usage Examples

Full Review

/pr-review 123

═══════════════════════════════════════════════════
PR Review Report: #123
═══════════════════════════════════════════════════

Summary:
  Files changed: 12
  Lines: +234 / -45

Findings: 8 total
  🔴 Critical: 1
  🟠 Major: 2
  🟡 Minor: 3
  💡 Suggestions: 2

[Detailed findings...]

VERDICT: REQUEST_CHANGES
═══════════════════════════════════════════════════

Filter Findings

/pr-findings 123 --category security

# Shows only security-related findings

Quick Summary

/pr-summary 123

# Shows change overview without full analysis

Output

Review reports include:

• Summary statistics
• Findings grouped by severity
• Code snippets with context
• Suggested fixes
• Overall verdict (APPROVE/COMMENT/REQUEST_CHANGES)

Verdict Logic

Condition	Verdict
Any critical finding	REQUEST_CHANGES
2+ major findings	REQUEST_CHANGES
Only minor/suggestions	COMMENT
No significant findings	APPROVE

Integration

For CLAUDE.md integration instructions, see claude-md-integration.md.

License

MIT