wshobson/agents
Multi-harness marketplace of plugins, subagents, skills, and commands for AI coding tools.
About
wshobson/agents is a large community collection of prompt components for AI coding tools: 94 plugins, 202 subagent definitions, 184 skills, and 105 commands, all authored in one Markdown source and adapted per harness via adapters. It works across Claude Code, OpenAI Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot, and Pi, covering Python/JS development, code review, testing, infrastructure, and security work. The repo is MIT-licensed, actively maintained (600+ commits since Jul 2025), and validates plugin structure in CI, with skill-only install paths for individual skills.
What it can do
- 94 installable plugins across Python/JS development, code review, testing, infrastructure, security
- 202 subagent definitions (code-reviewer, debugger, security-auditor, architect-reviewer, performance-engineer, and more)
- 184 skills: reference material agents load when relevant
- 105 commands: named workflows such as scaffolding and security scans
- One Markdown source per component, adapted to each harness by adapters (agents, model aliases, tool permissions)
- Generated artifacts for OpenCode, Antigravity CLI, GitHub Copilot, and Pi from the same source
- Structural validation and experimental static/plugin eval checks in CI
Setup
- In Claude Code, register the marketplace:
/plugin marketplace add wshobson/agents. - Install a plugin, e.g.
/plugin install python-development@claude-code-workflows(see the plugin catalog for all 94). - Ask Claude to use the plugin's agents, or invoke a command directly, e.g.
/python-development:python-scaffold Create a FastAPI service with tests. - For a single skill instead of a plugin:
gh skill install wshobson/agents python-testing-patterns(GitHub CLI 2.90+) ornpx skills add wshobson/agents --skill python-testing-patterns, selecting your coding tool. - For OpenCode, Antigravity CLI, Copilot, or Pi: clone the repo and run the harness installer, e.g.
make install-opencode(requires uv and Python 3.12+).
Prerequisites
Claude Code (or another supported harness); GitHub CLI 2.90+ or Node.js/npm for skill-only installs.
Use cases
Works with
Trust score — 9/10
One of the most-starred (40.5k) and actively maintained prompt-component collections: 94 plugins, 202 agents, 184 skills, 105 commands, with a real multi-harness story (Claude Code, Codex, Cursor, OpenCode, Antigravity, Copilot, Pi) and structural validation in CI. Caveats: quality inevitably varies across ~500 components; the repo's own README discloses that its LLM-judge/MC eval layers are experimental and "do not prove that a task succeeds in a project"; agent model aliases (opus/sonnet/haiku) need tuning to your budget. Metrics-v1 score 9/10 — adoption 8.6, maintenance 7.9, openness 10, manual 8.
★ 40,478 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.6/10
- Maintenance · recency, releases, issue load
- 7.9/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · 40,480 GitHub stars (log-scaled vs catalog peers)
- · last push 1d ago; 0 releases/12mo; 7 open issues
- · manual review: repo's own LLM-judge evals experimental/unvalidated; per-harness fidelity not fully checked
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.