Skip to content
Get packs

Guide · Open source

How to use agents: install wshobson/agents in your AI coding tool

Updated · written with AI from the project's own documentation and checked against our catalogue data

wshobson/agents is an open-source collection of plugins, specialist agents, skills and commands for AI coding tools. It works inside Claude Code, OpenAI Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot and Pi, so you need one of those tools already installed. This guide shows you how to pick a route, install a plugin or a single skill, and start using the agents and commands.

wshobson/agentswshobson/agents · ★ 42.1k · MIT

Quick start

  1. 1Pick your tool. Choose the harness you already use, such as Claude Code, Codex CLI, Cursor or Copilot.
  2. 2Install prerequisites. Install the requirements for your route: GitHub CLI 2.90 or later, Node.js and npm, or uv with Python 3.12 or later.
  3. 3Register the marketplace. In Claude Code, run the marketplace add command to register wshobson/agents.
  4. 4Install a plugin. Install a plugin such as python-development using the plugin install command.
  5. 5Invoke an agent or command. Ask your tool to use the relevant agent, or run a plugin command such as the scaffolding command.
  6. 6Install a single skill (optional). Use gh skill install or npx skills add to get one skill without the full plugin set.

Check what you need before installing

The components only run inside a supported AI coding tool, called a harness in the project's README. Choose the tool you already use and follow its setup route.

Prerequisites depend on the route you take. The skills-only route with gh needs GitHub CLI 2.90 or later. The npx route needs Node.js and npm. The clone-and-install route needs uv and Python 3.12 or later.

  • Claude Code, Codex CLI, Cursor, OpenCode, Copilot, Antigravity CLI or Pi, depending on your tool
  • GitHub CLI (gh 2.90 or later) for the gh skill route
  • Node.js and npm for the npx skills route
  • uv and Python 3.12 or later for the clone-and-install route

Install a plugin in Claude Code

In Claude Code, register the marketplace and then install a plugin. The README uses the Python development plugin as its example. Run these commands inside Claude Code, not in a terminal.

  • Plugin catalog: 94 plugins, covering Python, JavaScript and TypeScript, code review, testing, infrastructure and security
  • Pick the plugin that matches your work, such as python-development or security-scanning
/plugin marketplace add wshobson/agents /plugin install python-development@claude-code-workflows

Use an agent or a command

After installing, ask Claude to use the Python agents for your task. You can also run a plugin command directly. The README gives a scaffolding command as an example.

Agents are instructions for specialist subagents that handle delegated work. Skills are reference guidance an agent loads when relevant. Commands are named workflows you invoke yourself.

/python-development:python-scaffold Create a FastAPI service with tests

Install only one skill

If you do not want a full plugin, you can install a single skill. Both installers read skills from GitHub without cloning the repository. They install the skill without the plugin's agents, commands or hooks. Select your coding tool when prompted.

The project also documents other agents, install scopes and version pinning in its skills-only installation guide.

gh skill install wshobson/agents python-testing-patterns npx skills add wshobson/agents --skill python-testing-patterns

Install for OpenCode, Antigravity CLI, Copilot or Pi

For these four tools, clone the repository and run the installer for your tool. Each target generates the artifacts and links all local plugins into that tool's user configuration. Pi's agents need its reference subagent extension or a compatible one.

  • Clone with gh repo clone (GitHub CLI)
  • Run the make target for your tool
  • Pi agents require a subagent extension
gh repo clone wshobson/agents ~/agents cd ~/agents make install-opencode make install-antigravity make install-copilot make install-pi

Know the limits before you rely on it

The project's README lists no latest release, so it is not a fully stable, versioned product. Available components depend on the installation route and each harness's capabilities. For Codex CLI, native installs read the original skill bodies, while generated copies may split large bodies into reference files.

The plugin-eval tool gives static skill checks without model calls. Its LLM judge and Monte Carlo layers are experimental and not validated against human labels, so their scores do not prove a task will succeed.

  • Check which components your harness supports in the capability matrix before installing
  • Treat plugin-eval scores as model assessments, not proof of success
make generate-all make validate STRICT=1 make garden

What people use it for

  • Scaffolding a Python project or FastAPI service with tests using the python-development plugin
  • Adding a shared code review or security scanning workflow that a team can use across different tools
  • Installing only the python-testing-patterns skill for guidance on writing tests
  • Using JavaScript and TypeScript project setup guidance from the javascript-typescript plugin
  • Running a static skill check with plugin-eval before you change a skill

Frequently asked questions

Is wshobson/agents free to use?

The project is released under the MIT licence, as stated in its listing. The README does not mention prices. Model calls made by the optional evaluation tool use your ANTHROPIC_API_KEY when it is set, which bills that API account, or your Claude Code login otherwise.

What do I need installed to use wshobson/agents?

You need one supported AI coding tool: Claude Code, Codex CLI, Cursor, OpenCode, Copilot, Antigravity CLI or Pi. The components only run inside those tools. Clone-based installs also need uv and Python 3.12 or later.

Can I install just one skill instead of the whole plugin?

Yes. Use gh skill install wshobson/agents python-testing-patterns or npx skills add wshobson/agents --skill python-testing-patterns. These install the skill without the plugin's agents, commands or hooks.

Which AI coding tools does wshobson/agents support?

The README lists Claude Code, OpenAI Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot and Pi. Each tool has its own installation route, and available components vary by tool.

What is the difference between an agent, a skill and a command?

An agent is instructions for a specialist subagent that handles delegated work. A skill is reference guidance that an agent loads when relevant. A command is a named workflow you invoke yourself, such as project scaffolding or a security scan.

Are the plugin-eval scores reliable?

The README says the LLM judge and Monte Carlo layers are experimental and not validated against human labels. Their scores describe model assessments and repeated simulations, and they do not prove that a task will succeed in your project.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

wshobson/agents suits developers who already use a supported AI coding tool and want ready-made Python, JavaScript/TypeScript, review, testing or security agents, or who only need one skill. Skip it if you do not use any supported tool, or if you need a fully stable, versioned release, since the README lists no latest release.

Install options & downloads