Our take
- Activity
- 91
- Community
- 55
- Issues
- 95
- Pull requests
- 80
+57 stars in the last 10 days
- Releases more often than 93% of the projects we track
- Merges more pull requests than 87% of the projects we track
- Closes issues faster than 82% of the projects we track
Overview
GitHub Agentic Workflows, or gh-aw, lets developers describe repository tasks in Markdown with YAML frontmatter, then compile them into standard GitHub Actions workflows. It adds AI reasoning to repository automation while leaving deterministic builds, tests, and deployments to conventional CI.
The project is aimed at teams that want agents to interpret issues, pull requests, or CI results within an existing GitHub Actions setup. Agent jobs are read-only and sandboxed by default, with configured writes handled through separate safe-output jobs.
Key Features
- Defines workflow triggers, permissions, tools, and AI engine in YAML frontmatter, with instructions in a Markdown body.
- Compiles workflow source with
gh aw compileinto a generated.lock.ymlfor GitHub Actions. - Supports GitHub Copilot, Claude Code, OpenAI Codex, Google Gemini, and Pi.
- Uses read-only access and sandboxing as defaults for agent jobs.
- Buffers and validates configured GitHub writes through safe outputs and scoped permissions.
- Provides examples and documentation for workflow creation, engines, security, and common tasks.
Use Cases
- Repository maintainers can triage issues when classification requires interpreting free-form reports.
- Teams can add AI-assisted pull-request review alongside their existing deterministic checks.
- Developers can investigate CI failures that require synthesizing logs and repository context.
- Documentation owners can automate maintenance and reporting tasks that depend on understanding project content.
Getting Started
Install the GitHub CLI extension:
gh extension install github/gh-aw
See the README and quickstart guide for setup and configuration details.
Alternatives
- no-mistakes: no-mistakes focuses on validating and fixing Git pushes in an isolated AI pipeline, while gh-aw defines broader repository automations that run through GitHub Actions.
- aidevops: aidevops coordinates agents through an OpenCode plugin and shell framework, rather than defining repository-triggered automations in Markdown for GitHub Actions.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| gh-aw | Go | MIT | 5.3k | Not checked yet |
| no-mistakes | Go | MIT | 8.8k | Active |
| aidevops | Shell | MIT | 410 | Active |
Considerations
Agentic workflows are appropriate for tasks involving reasoning or interpretation, not as a replacement for reproducible scripts and ordinary CI. Safe defaults reduce risk but do not remove it: workflow authors should review permissions, tools, network access, and generated files, and supervise execution. The README also notes that releases from 0.83.3 through versions before 0.85.4 were retired after a security vulnerability; verify release guidance before installing or upgrading.
Found this useful?
Share it with someone who would like gh-aw.