ProblemClaw

AI-powered codebase scanner that identifies potential issues and creates GitHub issues automatically.

ProblemClaw uses Google Gemini to analyze your source code for bugs, security vulnerabilities, performance problems, and other issues — then creates well-structured GitHub issues for anything it finds.

Quick Start

# One-time setup — saves keys locally
npx problemclaw auth

# Scan current repo (auto-detects GitHub remote)
npx problemclaw scan --dry-run

# Create issues for real
npx problemclaw scan

Usage

Scan

problemclaw scan [options]

Options:
  -r, --repo <owner/repo>      GitHub repo (auto-detected from git remote)
  -d, --dry-run                 Preview issues without creating them
  -j, --json                    JSON output
  -s, --severity <level>        Min severity: critical|high|medium|low (default: medium)
  -c, --categories <list>       Comma-separated categories
  -m, --max-issues <n>          Max issues to create (default: 10)
  --model <model>               Gemini model (default: gemini-2.5-flash)

Auth

problemclaw auth    # Set up or check API keys (saved to ~/.config/problemclaw/)

Init

problemclaw init    # Create .problemclawrc.json with defaults

Configuration

Create .problemclawrc.json in your project root:

{
  "severity": "medium",
  "maxIssues": 10,
  "labels": ["problemclaw"],
  "exclude": [
    "**/node_modules/**",
    "**/dist/**",
    "**/build/**"
  ]
}

Authentication

ProblemClaw auto-detects GitHub credentials:

1. gh CLI (recommended): If you have gh installed and authenticated, it works automatically
2. Environment variable: GITHUB_TOKEN

For Gemini, run problemclaw auth once to save your API key.

Issue Categories

| Category | Description | |---|---| | bug | Logic errors, incorrect behavior | | security | Injection, auth bypasses, data exposure | | performance | N+1 queries, memory leaks, blocking ops | | maintainability | Dead code, excessive complexity | | error-handling | Swallowed errors, missing boundaries | | type-safety | Unsafe assertions, missing null checks | | concurrency | Race conditions, deadlocks | | memory-leak | Event listener leaks, unclosed resources |

Development

npm test              # Run all tests
npm run build         # Build with tsup
npm run lint          # Type-check

License

MIT