gh-aw: Define AI-powered repository automation in Markdown and run AI agents through GitHub Actions.
What problem it solves
Repository automation often needs reasoning or interpretation for tasks such as issue triage, pull-request review, CI failure investigation, documentation maintenance, dependency analysis, and repository reporting. The project provides a way to define these agentic workflows in Markdown with YAML frontmatter and compile them into standard GitHub Actions workflows while retaining explicit controls for permissions and writes.
Key features
Markdown workflow definitions — Workflow authors combine YAML frontmatter configuration with a Markdown body that describes the agent's task. Frontmatter controls triggers, permissions, tools, and the AI engine.
Source: README
Standard GitHub Actions compilation — The `gh aw compile` command validates workflow source and generates the `.lock.yml` workflow that GitHub Actions executes.
Source: README
Multiple built-in AI engines — Workflows can select among GitHub Copilot, Claude Code, OpenAI Codex, Google Gemini, and Pi as their AI engine.
Source: README
Sandboxed and read-only agent jobs — Agent jobs use read-only GitHub access and sandboxed execution by default. Configured writes are buffered, validated, and applied by separate `safe-outputs` jobs with scoped permissions.
Source: README
Complementary to conventional CI/CD — The project keeps deterministic builds, tests, linting, deployments, and reproducible scripts in conventional GitHub Actions, adding agentic workflows where interpretation is needed.
Source: README
Quick start
Copied verbatim from the project's README.
gh extension install github/gh-awWho it may suit
This may suit developers and repository maintainers who need AI-assisted issue triage, pull-request review, CI investigation, documentation maintenance, dependency analysis, or repository reporting through GitHub Actions.
Topics to discuss
Which repository automation tasks are most suitable for agentic workflows rather than conventional CI/CD jobs?
How should workflow authors balance scoped permissions and network access with agent capabilities?
Which AI engine selection and safe-output patterns work best for different kinds of repository automation?
Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.
Read the original README https://github.com/github/gh-aw#readme
ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.
Comments
…