The Rosetta Stone of AI coding agent plugins

What it does This marketplace packs 83 installable plugins—191 domain-specific agents, 155 skills, and 102 slash commands—into a single plugins/ directory. Claude Code consumes them natively, while the repo generates idiomatic, harness-native artifacts for OpenAI Codex CLI, Cursor, OpenCode, Gemini CLI, and GitHub Copilot. Installing a plugin loads only its components into context, not the entire catalog.

The interesting bit Instead of lowest-common-denominator translations, the build pipeline emits target-specific artifacts—TOML for Gemini, Markdown profiles for Copilot, Cursor rules—from the same Markdown source. It also tiers model usage by task criticality (architecture gets Opus 4.7, docs get Sonnet, quick ops get Haiku) and ships a three-layer evaluation framework that can statically lint, LLM-judge, or Monte-Carlo-test any plugin.

Key highlights

• 83 plugins, 191 agents, 155 skills, 102 commands, and 16 multi-agent orchestrators
• One source of truth in plugins/, auto-discovered from directory structure
• Native artifacts for five harnesses: Claude Code, Codex CLI, Cursor, OpenCode, Gemini CLI, and Copilot
• plugin-eval quality framework: static analysis, LLM judge, and Monte Carlo certification
• Progressive-disclosure skills that load only when activated

Caveats

• Per-harness “gotchas” are documented; Gemini and OpenCode require clone-and-generate workflows rather than direct marketplace installs
• Drift, dead links, and harness-specific cap violations are expected, so the repo includes dedicated detection tooling

Verdict Worth bookmarking if you live in Claude Code, Cursor, or Codex and want pre-built domain expertise. Skip it if you don’t use agentic IDEs or prefer hand-rolled prompts.