General OAT Skills
A curated collection of skills for Claude Code, organized as a plugin using the Agent Skills open standard.
Install
As a Plugin (Recommended)
From inside Claude Code, first add the marketplace:
/plugin marketplace add Open-Agent-Tools/General-OAT-Skills
Then install at the level you need:
Everything — all 10 skills:
/plugin install general-oat-skills@general-oat-skills
Bundles — thematic groups:
/plugin install general-oat-skills@oat-core # check, cleanup, test, load, publish-python, publish-rust
/plugin install general-oat-skills@oat-quality # qa, teach-me
/plugin install general-oat-skills@oat-agents # scaffold, run-adk-evals
Individual skills — pick exactly what you need:
/plugin install general-oat-skills@oat-check
/plugin install general-oat-skills@oat-cleanup
/plugin install general-oat-skills@oat-test
/plugin install general-oat-skills@oat-load
/plugin install general-oat-skills@oat-publish-python
/plugin install general-oat-skills@oat-publish-rust
/plugin install general-oat-skills@oat-qa
/plugin install general-oat-skills@oat-scaffold
/plugin install general-oat-skills@oat-teach-me
/plugin install general-oat-skills@oat-run-adk-evals
Manual
Clone and symlink into your project:
git clone https://github.com/Open-Agent-Tools/General-OAT-Skills.git
ln -s "$(pwd)/General-OAT-Skills/skills" /path/to/your-project/.claude/skills
Or copy individual skill directories into .claude/skills/ (project-level) or ~/.claude/skills/ (global).
Skills
Core Dev Workflow
| Skill | Description |
|---|
/check | Quick project health check — git status, recent commits, GitHub Actions, package build |
/cleanup | Full code quality pipeline — ruff, mypy, pytest — with validation loops |
/test | Run pytest suite with coverage reporting and auto-fix on failures |
/load [directory] | Load all markdown files from a directory into context |
/publish-python <version> | Python release — cleanup, version bump, uv build, GitHub release |
/publish-rust <bump> | Rust crate/workspace release — verify gates, version bump, signed tag, CI or direct crates.io publish |
Quality & Learning
| Skill | Description |
|---|
/qa [focus] | Multi-agent QA audit — code quality, architecture, security, tests, CI/CD, docs |
/teach-me <topic> [flags] | Interactive lesson-based teaching with web research, quizzes, and difficulty levels |
AI Agent Development
| Skill | Description |
|---|
/scaffold <template> [name] | Create a new project from 18 templates with full setup |
/run-adk-evals <folder> | Run Google ADK evaluations with detailed results and rate limiting |
Scaffold Templates
General Purpose
python-lib python-cli python-web python-data node-lib node-web rust-lib go-cli
AI Agent Frameworks
python-langchain python-adk python-crewai python-autogen python-swarm python-phidata python-llama-index python-haystack python-semantic-kernel python-agency-swarm
Skill Behaviors
- Side-effect protection —
cleanup, publish-python, publish-rust, scaffold, and qa have disable-model-invocation: true, so Claude won't auto-trigger them. They must be explicitly invoked.
- Context isolation —
load and check run in forked context (context: fork) to avoid polluting your main conversation.
- Argument hints — Skills that accept parameters show usage hints in autocomplete.
Project Structure
General-OAT-Skills/
├── .claude-plugin/
│ ├── marketplace.json # all install targets (full, bundles, individual)
│ └── plugin.json
├── skills/ # canonical skill definitions
│ ├── check/SKILL.md
│ ├── cleanup/SKILL.md
│ ├── load/SKILL.md
│ ├── publish-python/SKILL.md
│ ├── publish-rust/SKILL.md
│ ├── qa/SKILL.md
│ ├── run-adk-evals/SKILL.md
│ ├── teach-me/SKILL.md
│ ├── test/SKILL.md
│ └── scaffold/
│ ├── SKILL.md
│ └── templates/ # 18 project templates
├── skills-core/ # bundle: symlinks → check, cleanup, test, load, publish-python, publish-rust
├── skills-quality/ # bundle: symlinks → qa, teach-me
├── skills-agents/ # bundle: symlinks → scaffold, run-adk-evals
├── skill-*/ # individual: one symlink each (skill-check/, skill-qa/, etc.)
├── CODE_OF_CONDUCT.txt
├── CONTRIBUTING.txt
├── LICENSE
└── SECURITY.txt
Contributing
See CONTRIBUTING.txt for full guidelines.
To add a new skill:
- Create
skills/<skill-name>/SKILL.md with YAML frontmatter (name, description, allowed-tools)
- Test locally by symlinking the repo's
skills/ directory into a project
- Submit a pull request
License
Apache License 2.0 — see LICENSE.
