github-pr-analyzer

name	github-pr-analyzer
description	Use when the task involves analyzing pull requests in a GitHub repository for staleness, age distribution, or identifying PRs that need attention. Fetches all open PRs with pagination and generates comprehensive reports with actionable recommendations.

GitHub PR Analyzer

Analyzes pull requests in any GitHub repository to identify stale PRs, age distribution, and provide actionable recommendations for PR triage.

When to Use

Use this skill when you need to:

• Identify stale or abandoned pull requests (30+ days without updates)
• Analyze PR age distribution across different staleness buckets
• Generate reports for PR triage sessions
• Find PRs that need review attention
• Create actionable recommendations for maintainers

Instructions

Basic Usage

import { analyzeStalePRs } from './.claude/skills/github-pr-analyzer/implementation';

const result = await analyzeStalePRs('owner', 'repo', {
  outputDir: './workspace',
  staleDays: 30
});

console.log(`Total open PRs: ${result.stats.totalOpen}`);
console.log(`Stale PRs: ${result.stats.staleCount}`);
console.log(`Report saved to: ${result.reportPath}`);

Options

• outputDir: Directory to save report and raw data (default: ./workspace)
• staleDays: Number of days to consider a PR stale (default: 30)

Output Files

The skill generates two files:

1. stale-pr-report.md: Markdown report with:

   • Summary statistics
   • Age distribution table
   • Top 10 most stale PRs
   • Recommended next actions

2. open-prs.json: Raw PR data for custom analysis

Report Structure

The generated report includes:

• Summary: Total open PRs and stale PR count with percentage
• Age Distribution: Breakdown into buckets:
  • Fresh (< 30 days)
  • Stale (30–60 days)
  • Very stale (60–90 days)
  • Abandoned-looking (> 90 days)
• Top Stale PRs: 10 PRs with longest time since last update
• Recommendations: Prioritized action items for triage

Examples

Analyze with Custom Staleness Threshold

// Consider PRs stale after 14 days
const result = await analyzeStalePRs('anthropics', 'anthropic-sdk-python', {
  staleDays: 14
});

Custom Output Directory

const result = await analyzeStalePRs('owner', 'repo', {
  outputDir: './reports/pr-analysis'
});

Gotchas

• Pagination: The skill automatically handles pagination, fetching up to 100 PRs per page. This works for repositories with thousands of PRs.
• Rate Limiting: GitHub API has rate limits. For large repositories (1000+ PRs), consider authenticating to get higher rate limits.
• Draft PRs: Draft PRs are included in the analysis and marked in the report. You may want to filter these differently in custom analysis.
• Updated vs. Created: Staleness is based on updated_at, not created_at. A PR can be old but recently active.
• Working Directory: The script must be run from the project root (where .mcp.json exists), not from subdirectories.

Dependencies

• MCP Tools: list_pull_requests from the github server
• Node.js: Built-in fs/promises for file operations
• TypeScript: Uses tsx for execution

Related Skills

• github-issue-analyzer: Analyze repository issues instead of PRs
• comprehensive-issue-analyzer: Deep analysis of issues with pagination