driftcode-auditor

Local-first CLI for auditing AI-generated code — catch common mistakes in Cursor, Claude, Copilot, and other AI coding output before they reach production.

License: MIT — Free to use, modify, and distribute.

Features

• Automatically respects .gitignore and common build directories (.next, node_modules, dist, .git, etc.)
• Parallel scanning for good performance on large codebases
• Reports files scanned + time taken
• --stream mode to print issues as they are discovered
• --quiet and --verbose output modes
• Customizable file extensions via --ext or config file
• Configurable PII allowlist to reduce false positives
• Pure local operation — no network calls

Configuration

• Project-level: .driftcode.json in the scanned directory
• User-level: ~/.config/driftcode/config.json

Installation

Recommended (avoids system Python conflicts on modern distributions):

pipx install driftcode-auditor

Or with a virtual environment:

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install driftcode-auditor

Then run:

driftcode-auditor --help

Windows users: If driftcode-auditor is not found after pip install, use the module form instead:

python -m driftcode_auditor --help

Troubleshooting: externally-managed-environment error

This error occurs on PEP 668 systems (Ubuntu 23.04+, Debian 12+, etc.). Use pipx or an explicit virtual environment as shown above.

Usage

Basic scan:

driftcode-auditor --path /path/to/project --format md --privacy

Or run directly from source:

python -m driftcode_auditor --path /path/to/project --format md --privacy

Common options:

driftcode-auditor --path . --format md --privacy --maintainability
driftcode-auditor --path . --format json --privacy --stream
driftcode-auditor --path . --privacy --verbose
driftcode-auditor --path . --privacy --ext .py,.ts

Flags

Flag	Description
--path	Directory to scan (default: current dir)
--format	Output format: md or json
--privacy	Enable privacy risk detection
--maintainability	Enable maintainability checks
--stream	Print issues immediately as found
--output	File for full detailed report (default: driftcode-report.md)
--quiet	Minimal output
--verbose	Show every file being scanned
--ext	Comma-separated list of extensions to scan

Example Output

Scanning /home/user/project ... (skipping common build dirs)

# DriftCode Auditor Report
...

## Privacy
- **pii** in `lib/network.ts:42`: PII in code
- ... and 47 more

## Architecture
- **large_file** in `lib/commands.ts:1`: File >500 lines
- ... and 3 more

Scan complete.
Files scanned: 1247
Time taken: 3.21s
Total issues found: 87

When many issues are found, only the first 15 per category are shown, followed by a summary count.

Running Tests

python -m pytest tests/

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome — especially new rules, documentation improvements, and bug reports.

• See GitHub Discussions for ideas and feedback.
• Look for issues labeled good first issue or help wanted.
• For larger changes, please open an issue first.

We especially appreciate contributions from people who use AI coding tools daily (Cursor, Claude, Copilot, etc.) and want to help catch more real-world mistakes.

Support the Project

DriftCode Auditor is free and open source, and will remain that way.

If you find it valuable — especially when auditing AI-generated code — consider sponsoring its development on GitHub Sponsors.

Pro sponsors receive a manual invitation to a private repository containing advanced AI rules, priority support, and early access to new features.

Your support helps sustain focused work on the tool, funds new rule development, and keeps the project moving forward as AI coding practices evolve.

Every sponsor directly contributes to making reliable AI code review more accessible.

See CHANGELOG.md for release history.

Philosophy

DriftCode Auditor is designed to help developers audit AI-generated code for common and obvious mistakes.

Key goals:

• Catch simple issues that would otherwise go unnoticed
• Be fast and respectful of existing project structure (respects .gitignore)
• Stay simple, local-first, and privacy-conscious
• Provide clear, actionable feedback without false security

All scanning happens on your machine with no data leaving your environment.

Reviewing AI-Generated Code

Quickly audit AI-generated code from Cursor, Claude, Copilot, Windsurf and other AI coding tools.

A short terminal demo / asciicast will be added here shortly (recording in progress).

DriftCode Auditor is designed to help you quickly audit code produced by AI coding assistants (Claude, Cursor, Copilot, etc.).

Common Issues It Catches

• Missing or fake error handling
• Hardcoded secrets and credentials
• Overly generic function names
• Architectural drift and pattern violations
• Missing edge case handling
• Potential PII / secret leaks

Recommended Workflow

1. Generate code with your AI assistant
2. Run driftcode-auditor on the changed files
3. Review flagged issues before committing
4. Use --output report.md for a full detailed report

Example: Catching a Common AI Mistake

AI-generated code:

def get_user(user_id):
    return db.query("SELECT * FROM users WHERE id = " + user_id)

DriftCode Auditor output:

- **pii** in `user_service.py:12`: PII in code → `return db.query("SELECT * FROM users WHERE id = " + user_id)`
- **secret** in `user_service.py:12`: Potential SQL injection risk

This helps catch problems that are easy to miss in normal code review.

This adds a lightweight but effective safety net when working with AI-generated code.

DriftCode Auditor Pro (Paid)

Pro features are delivered via a separate private repository for sponsors (see PAID_OFFERING.md). The free CLI has no access to paid features.