Agent Toolkit

Portable Agent Skills for coding agents. This repo is skills-only for now and uses the Agent Skills SKILL.md format so skills can be installed for Codex, Claude Code, GitHub Copilot, and other supported agents.

Install

Install the Terraform style skill for one agent:

gh skill install vegardx/agent-toolkit terraform-style --agent codex --scope user
gh skill install vegardx/agent-toolkit terraform-style --agent claude-code --scope user
gh skill install vegardx/agent-toolkit terraform-style --agent github-copilot --scope user

Install every skill in the repo:

gh skill install vegardx/agent-toolkit --all --agent codex --scope user

Pin to a tag or commit when repeatability matters:

gh skill install vegardx/agent-toolkit terraform-style@v0.1.0 --agent codex --scope user

Skills

User-invoked

• skill-authoring — Design, write, review, evaluate, and retire predictable Agent Skills. Invoke it explicitly with /skill-authoring.

Model-invoked

• github — GitHub platform behavior across GitHub.com, GHE.com Data Residency, and GitHub Enterprise Server.
• github-agentic-workflows — GitHub Agentic Workflows, gh aw, permissions, network policy, safe outputs, and validation.
• github-cli — GitHub CLI commands, auth, host and repository targeting, APIs, output, and safe mutation.
• terraform-style — Terraform and HCL conventions for modules, variables, composition, state, documentation, and tests.

Future skills should be added only after their conventions have been deliberately worked through. Avoid checking in generated placeholder skills.

Layout

agent-toolkit/
  README.md
  LICENSE
  AGENTS.md

  skills/
    github/
      SKILL.md
      references/
        official-docs.md

    github-agentic-workflows/
      SKILL.md
      references/
        official-docs.md
        examples.md

    github-cli/
      SKILL.md
      references/
        examples.md

    skill-authoring/
      SKILL.md
      agents/
        openai.yaml
      evals/
        cases.json
      references/
        evaluation.md
        structure.md
        taxonomy.md

    terraform-style/
      SKILL.md
      references/
        module-patterns.md

Canonical skills live under skills/. Only add optional directories a skill actually uses:

skills/
  <skill-name>/
    SKILL.md
    agents/       # harness metadata, when needed
    references/   # selectively loaded guidance
    scripts/      # deterministic mechanics
    assets/       # templates and static inputs
    evals/        # behavior cases and fixtures

Plugins, hooks, agents, MCP servers, and other packaging layers may be added later, but they are not part of the current skills-only distribution model.

Publish

Validate before publishing:

gh skill publish --dry-run

Publish a release:

gh skill publish --tag v0.1.0

Publishing validates discovered skills and creates a GitHub release for the selected tag.

License

This repository is licensed under the MIT License. See LICENSE.