wshobson/agents 🤖 Agent Free

What it does

wshobson/agents is a plugin marketplace of ready-made building blocks for agentic coding. It started as a Claude Code subagent collection and has grown into a catalogue of 94 plugins that together hold roughly 200 specialist agents, over 180 Agent Skills, more than 100 slash commands and a set of multi-agent orchestration workflows. The specialists cover architecture, individual languages and frameworks, infrastructure, security, data and machine learning, documentation, business and SEO.

What is inside / how it works

Every plugin is a small, single-purpose unit. The python-development plugin, for example, bundles a few Python agents, a scaffolding command and a group of Python skills. Installing a plugin loads only its own components into context, not the whole marketplace, so you can keep sessions lean by adding just what a project needs. Agents are assigned model tiers, with heavier reasoning work such as architecture and security review mapped to Opus and quicker operational jobs to Sonnet or Haiku. A bundled evaluation framework, plugin-eval, scores skill quality with static checks, an LLM judge and repeated simulated runs.

Works with

Claude Code is the native target and source of truth. From the same Markdown source the repo generates harness-specific output for OpenAI Codex CLI, Cursor, OpenCode, the Antigravity CLI, GitHub Copilot and Pi. Skills on their own can also be installed into other agents with gh skill or npx skills.

How to install or start

In Claude Code, add the marketplace and then install individual plugins:

/plugin marketplace add wshobson/agents
/plugin install python-development

Codex and Cursor read committed registries; OpenCode, Antigravity and Pi install by cloning the repo and running the documented make targets. See the repo's docs/harnesses.md for per-tool steps.

Pricing

Free and open source under the MIT licence. You only pay for the model usage of the agent you run it in.

Who should use it

Developers who want a broad library of specialists without writing their own, and teams that use more than one coding agent. The main risk is installing too many plugins at once, which adds context and makes it harder to know which agent is doing what.