Our take
Activeregular commits and releases- Activity
- 90
- Community
- 56
- Issues
- 95
- Pull requests
- 78
+91 stars in the last 16 days
- Releases more often than 93% of the projects we track
- Merges more pull requests than 84% of the projects we track
- Closes issues faster than 82% of the projects we track
This is an actively maintained, broadly contributed tool for adding agent-driven tasks to GitHub Actions, with meaningful security controls but a need for careful workflow review.
Good fit if
- Your team uses GitHub Actions and has repository tasks that require reasoning rather than deterministic scripts.
- You want to author those tasks in Markdown and use one of the supported AI engines.
- You can review workflow permissions and configured writes; the project is MIT-licensed and its development shows regular commits and releases.
Look elsewhere if
- You need conventional CI automation only, since the project is designed to complement rather than replace deterministic workflows.
- Your automation platform is not GitHub Actions or you do not want to install and use a GitHub CLI extension.
- Your team cannot supervise agent behavior and carefully review security settings and generated workflow files.
All health signals
| Last commit | 2026-10-23 (today) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 100+ (latest v0.93.1, 2026-10-23, pre-release) |
| Contributors | 40 (top contributor: 69% of commits) |
| Issues closed, last 90 days | 65+ (typically closed in 0 days) |
| Pull requests merged, last 90 days | 74 (typically merged in 0 days) |
| Project age | 15 months |
Checked on 2026-10-23 with the GitHub API.
Overview
GitHub Agentic Workflows (gh-aw) is a Go-based GitHub CLI extension that turns Markdown workflow definitions into standard GitHub Actions workflows. It combines YAML frontmatter for triggers, permissions, tools, and engine settings with a Markdown brief describing the task, letting teams add AI reasoning to repository automation that is difficult to express as fixed scripts.
It is aimed at developers already using GitHub Actions who want agents to handle interpretive work, such as triage or investigating failures, without replacing deterministic build, test, and deployment pipelines. The project supports GitHub Copilot, Claude Code, OpenAI Codex, Google Gemini, and Pi as AI engines.
Key Features
- Define agentic workflows in Markdown with YAML frontmatter.
- Compile definitions with
gh aw compileinto lock files that GitHub Actions can execute. - Configure workflow triggers, permissions, tools, and AI engines.
- Use supported AI engines including GitHub Copilot, Claude Code, OpenAI Codex, Google Gemini, and Pi.
- Run agent jobs read-only and sandboxed by default.
- Route configured repository writes through validated
safe-outputsjobs with scoped permissions. - Find workflow examples and setup guidance in the project documentation.
Use Cases
- Repository maintainers can use an agent to label and triage incoming issues when interpretation is needed.
- Reviewers can automate an initial pull-request review while keeping normal review and merge controls in place.
- CI owners can ask an agent to investigate failed runs and summarize likely causes.
- Documentation teams can automate routine maintenance that requires understanding repository context.
- Engineering teams can generate dependency analyses or repository reports on a GitHub Actions schedule.
What you need
Detected in the repository
- Go 1.27.3 or newer (from go.mod)
- A Dockerfile, so it can run in a container
- A test suite and automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install the extension:
gh extension install github/gh-aw
Then follow the quickstart to choose an engine, add a sample workflow, and run it through GitHub Actions. See the README for setup and project details.
Alternatives
- run-gemini-cli: It runs Gemini CLI directly in GitHub Actions, while gh-aw defines Markdown workflows and compiles them for supported agents.
- actions: It runs Outfitter profiles as agent jobs in GitHub Actions, while gh-aw provides a Markdown-based workflow authoring and compilation layer.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| gh-aw | Go | MIT | 5.4k | Active |
| run-gemini-cli | TypeScript | Apache-2.0 | 2.1k | Active |
| actions | Shell | MIT | 2 | New |
Considerations
Agent output is not deterministic in the way ordinary build and test scripts are, so workflows need human oversight. Although agent jobs are read-only and sandboxed by default, permissions, tools, network access, safe outputs, and generated files remain configurable and require review. The project is actively maintained and has contributions from multiple contributors, but its security-sensitive workflow model merits particular care. The README also reports that releases from versions 0.83.3 through versions before 0.85.4 were retired after a security vulnerability; check the advisory and use a current release. Running workflows requires GitHub Actions and the Go-based CLI extension, and AI engine selection and authentication are part of setup.
Found this useful?
Share it with someone who would like gh-aw.