codex-action  by openai

GitHub Actions integration for secure AI code analysis

Created 1 month ago
651 stars

Top 51.3% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

This GitHub Action enables running the OpenAI Codex CLI directly within GitHub Actions workflows, offering fine-grained control over Codex's execution privileges. It's designed for developers seeking to automate code analysis, review, or generation tasks as part of their CI/CD pipelines, providing a secure and customizable integration.

How It Works

The action installs the Codex CLI and sets up a secure proxy to the OpenAI API. It requires an OPENAI_API_KEY secret and integrates with standard GitHub Actions checkout steps. A key feature is the safety-strategy input, which allows users to restrict Codex's access to the runner environment, ranging from read-only to full access, enhancing security.

Quick Start & Requirements

  • Primary Command: Use uses: openai/codex-action@v1 in your GitHub Actions workflow YAML.
  • Prerequisites:
    • OPENAI_API_KEY stored as a GitHub Actions secret.
    • actions/checkout@v5 must be run prior to this action.
  • Dependencies: None beyond the action and API key.
  • OS Support: Linux/macOS support all safety-strategy options. Windows runners require safety-strategy: unsafe.
  • Links: Refer to security.md for detailed security practices (not provided).

Highlighted Details

  • Privilege Control: Robust safety-strategy options (drop-sudo, unprivileged-user, read-only, unsafe) to mitigate security risks.
  • Customization: Supports inline prompts or prompt files, allowing tailored code analysis tasks.
  • Workflow Integration: Designed to seamlessly integrate into CI/CD, with examples showing PR comment generation.
  • CLI Access: Can install and configure the Codex CLI for subsequent use within the same job.

Maintenance & Community

No specific details on contributors, community channels, or roadmap were provided in the README snippet.

Licensing & Compatibility

  • License: Apache License 2.0.
  • Compatibility: Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

Windows runners have limited safety-strategy options, defaulting to unsafe. The drop-sudo strategy impacts subsequent steps in the same job that require superuser privileges. Network access is disabled by default in the sandbox, necessitating pre-downloading dependencies for code execution tasks. Careful management of the OPENAI_API_KEY and selection of safety-strategy are critical for security.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
20
Issues (30d)
11
Star History
642 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems").

codegate by stacklok

0%
703
AI agent security and management tool
Created 11 months ago
Updated 5 months ago
Feedback? Help us improve.