39.5k stars · MIT · no tagged releases; 546 commits on main · Track this in Rising ↗
A marketplace of ready-made Claude Code plugins, subagents and skills — a free capability upgrade to how he builds everything else.
Radar check
What we found when we opened it. Every bar is one fact, with its date.
Last code pushed 26 Oct 2026, one day before this check.
No releases published on GitHub; the code date above is the one to trust.
Public since July 2025, a little over 1 year.
41,875 stars and 4,436 forks.
90 commits and 39 new issues in the last 90 days.
91 contributors. 79 pull requests merged in 90 days.
No security policy file says how to report a problem. No published security advisories.
How the bars are scored
- Alive. Days since the last code push: a week or less 100, a month 90, three months 70, six months 50, a year 30, eighteen months 15, older 0.
- Releases. Age of the latest GitHub release, on a similar scale; no bar when the project publishes no releases.
- Maturity. Age of the project: eight years 100, five 90, three 70, two 55, one 40, six months 25, younger 10.
- Adoption. Stars on a log scale: 100,000 stars is 100, 10,000 is 80, 1,000 is 60, 100 is 40.
- Activity. Commits in the last 90 days: 300 or more 100, 100 85, 30 65, 10 45, any 25.
- Community. People who have contributed: 100 or more 100, 30 80, 10 60, 5 40, 2 25, one 10.
- Security. A published security policy 80, none 40; published advisories are listed, not scored.
Checked by Repo Radar on 27 Oct 2026 from the repository on GitHub. Covered in Edition 4.
▶Repo detailsthe review · specs · pros & cons · install
What it does
It collects ready-made plugins, agents, skills and commands for coding assistants. At the time of this check its own documentation lists 93 plugins, 202 agents, 181 skills and 105 commands. Code review, deployment, security checking, documentation, testing and database work are all covered. Each plugin installs with a single command, with nothing to host and nothing to back up. It is no longer tied to one assistant: Claude Code, OpenAI Codex CLI, Cursor, OpenCode, Antigravity CLI and GitHub Copilot are all supported. It publishes no tagged releases, so an install takes whatever the main branch holds that day. It is a marketplace and not a curated set. Nothing filters overlap or quality, and it does not change the assistant itself.
Why it matters
Who it suits. Anyone who builds through a coding assistant and has never spent any time improving how it behaves. That is usually the highest-value hour available, because it improves every other task rather than one of them. Skip it if the current setup already works well; replacing a working configuration with thirty untested plugins makes things worse.
Verdict. Read three, install two, and delete anything unused after a fortnight. The realistic value is narrow and real: a couple of these, particularly around code review and deployment, save an evening each. Two cautions. Quantity has been chosen over curation, so many entries overlap and some are thin. And there are no tagged releases, so there is no version to pin and the contents will move under you. awesome-claude-code is the curated directory to read alongside it, at 53.1k stars. The support for six different assistants is new since this edition first ran, when the collection covered one.
Summary (Key Points)from the README
Key features
- Seven harnesses: One Markdown source serves Claude Code, OpenAI Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot, and Pi, with adapters generating each harness's format.
- Large catalog: 94 plugins, including 92 local plugins and 2 external entries.
- Component counts: The local source contains 202 agents, 185 skills, and 105 commands.
- Focused plugins: python-development covers Python, Django, FastAPI, testing, and packaging, and contains three agents, sixteen skills, and a scaffolding command.
- Skills-only install: Individual skills can be installed through GitHub CLI's `gh skill` command or the `npx skills` installer without cloning the repository.
- Model aliases: Source agents use opus, sonnet, haiku, fable, or inherit, and adapters map these to each harness's model settings.
- Automated checks: The repository validates plugin structure, generated artifacts, and documentation, and plugin-eval runs static skill checks without model calls.
- Optional evaluation layers: An LLM judge and a Monte Carlo layer score skills, requiring the optional llm dependencies.
Ways to install
| Platform or method | What it needs | Effort |
|---|---|---|
| Claude Code plugin marketplace | Commands run inside Claude Code to register the marketplace and install a plugin | Low: in-app commands only |
| OpenAI Codex CLI | A current Codex CLI in a terminal; native marketplace or generated artifacts | Low to medium: choose between native and generated routes |
| Cursor | Cursor and the committed plugin marketplace registry | Low: follow the Cursor setup instructions |
| OpenCode | Clone the repository, install uv and Python 3.12 or later, run `make install-opencode` | Medium: clone and build step |
| Antigravity CLI | Clone the repository, install uv and Python 3.12 or later, run `make install-antigravity` | Medium: clone and build step |
| GitHub Copilot | Clone the repository, install uv and Python 3.12 or later, run `make install-copilot` | Medium: clone and build step |
| Pi | Clone the repository, install uv and Python 3.12 or later, run `make install-pi`; agents need a subagent extension | Medium to high: clone, build, and extension setup |
| Skills only (`gh skill` or `npx skills`) | GitHub CLI 2.90 or later, or Node.js and npm | Low: single command per skill |
Important
- Partial component sets: Available components depend on the installation route and the harness's capabilities; the skills-only route omits agents, commands, and hooks.
- Codex size limit: Generated Codex copies split oversized skill bodies into reference files to fit an 8 KB budget, so they differ from the native source skills.
- Pi agents: Pi's agents require its reference subagent extension or a compatible one.
- Experimental evaluation: The LLM judge and Monte Carlo layers are not validated against human labels, and their scores do not prove a task succeeds in a project.
- API billing: Evaluation model calls use ANTHROPIC_API_KEY when set and bill that account; otherwise they use your Claude Code login.
- Pensyve discontinued hosted service: Pensyve Cloud closed on October 1, 2026, so its dashboard, API, and hosted MCP endpoints are unavailable; only the self-hosted runtime remains.
- Pensyve is external: It is an optional marketplace entry maintained by Major7 Apps, which the repository's maintainer founded.
- Mixed licensing: The repository is MIT licensed, but external plugins have their own licenses.
Written by an AI model from the project’s README; not part of the review, and it can be wrong.
- MIT, and installing a plugin is one command, with nothing to host or back up
- Improves the tool used to build everything else on this page, so one hour goes a long way
- Now works with six different coding assistants rather than one
- Quantity over curation: many entries overlap and some are thin
- No tagged releases, so an install takes whatever the main branch holds that day
- More installed plugins makes a setup more confusing, so discipline beats coverage
Instead of re-describing what a careful code reviewer or a deployment checklist should do every time you need one, you install a version someone has already written and iterated on.
/plugin marketplace add wshobson/agents /plugin install python-development
