Agents

This repository is a plugin marketplace built for Claude Code and generated for four other agentic harnesses (Codex CLI, Cursor, OpenCode, Antigravity CLI, and GitHub Copilot) from a single Markdown source. It's for developers who use AI coding agents and want a curated set of domain-expert agents, reusable skills, and slash commands rather than building prompts from scratch. It targets teams standardizing agentic workflows across languages, infrastructure, security, data, and ML tasks.

Key features

• 93 plugins: granular, single-purpose installable units covering languages, frameworks, infra, security, and business tasks; installing one loads only its components into context.
• 202 agents: domain experts spanning architecture, languages, infrastructure, security, data, ML, docs, business, and SEO.
• 181 skills: modular knowledge packages that use progressive disclosure, loading only when activated.
• 105 commands: slash commands for scaffolding, security scans, test generation, and infrastructure setup.
• 16 orchestrators: multi-agent coordination workflows for full-stack development, security review, ML pipelines, and incident response.
• Multi-harness generation: adapters emit harness-native artifacts (not generic translations) for Codex CLI, Cursor, OpenCode, Antigravity CLI, and Copilot from the same source.
• Tiered model strategy: agents map to model tiers (Opus, Sonnet, Haiku, inherit, or a premium long-horizon tier) based on task complexity and cost.
• PluginEval framework: a three-layer quality check (static analysis, LLM judge, Monte Carlo simulation) for scoring and certifying plugin or skill quality.
• External integrations: optional git-subdir entries for Pensyve (external memory) and HOL Guard (security scanning), installed as discrete plugins.

Ideal use cases

• You use Claude Code daily and want a ready-made catalog of agents and skills instead of writing custom prompts for every language, framework, or infra task.
• You run multiple agentic CLIs (Codex, Cursor, OpenCode, Antigravity, Copilot) and want a single source that generates harness-native plugin registries for each, instead of maintaining separate prompt libraries per tool.
• You're building or auditing agent skills and want a structured evaluation framework (static checks, LLM judging, Monte Carlo runs) to certify quality before shipping.
• You need multi-agent orchestration for complex work like large migrations, security reviews, or incident response, where a single agent isn't enough.
• It's not a fit if you don't use any of the supported harnesses, or if you want a minimal, single-agent setup rather than a large catalog of 200+ agents and 180+ skills to sift through.
• It's also not the right choice if you need agent behavior fully portable to arbitrary third-party tools outside the five supported harnesses; the project explicitly favors harness-native artifacts over lowest-common-denominator translations.

Installation

For Claude Code, add the marketplace and install individual plugins:

/plugin marketplace add wshobson/agents
/plugin install python-development          # or any of 93 plugins

For Codex CLI and Cursor, install natively from the committed registries:

npx codex-marketplace add wshobson/agents        # Codex; then install individual plugins
# Cursor: add the marketplace, then `/plugin install <name>`

For Antigravity CLI and OpenCode, clone the repo and generate the harness-specific trees:

gh repo clone wshobson/agents ~/agents && cd ~/agents
make generate HARNESS=antigravity && make install-antigravity  # Antigravity (agy)
make install-opencode                                          # OpenCode (runs generate + symlinks)

To regenerate all harness artifacts and run structural checks locally:

make generate-all                        # all five harnesses
make validate                            # structural checks
make garden                              # drift / dead-link / cap detection

To run the plugin quality evaluation tool (requires uv):

uv run plugin-eval score path/to/skill --depth quick
uv run plugin-eval certify path/to/skill

Full setup details, troubleshooting, and the per-harness capability matrix live in docs/usage.md and docs/harnesses.md.