Plugins & skills for AI coding agents

Guides a coding agent through the whole loop — brainstorm a spec, break it into bite-sized tasks, build with red-green-refactor TDD, then review and ship — and keeps it on track autonomously for hours.

/plugin install superpowers@claude-plugins-official

A cross-harness operating layer — 67 specialized agents, 277 skills, memory and security baked in — that runs the same setup across Claude Code, Cursor, OpenCode, Codex, Copilot, Gemini and Zed.

/plugin marketplace add https://github.com/affaan-m/ECC
/plugin install ecc@ecc

A big, well-organized marketplace of agents and plugins that works across Claude Code, Codex CLI, Cursor, OpenCode and Copilot — install once and pull in production-grade specialists.

/plugin marketplace add wshobson/agents

An interactive CLI that scaffolds Claude Code setups — agents, commands, hooks and CLAUDE.md memory — from ready-made templates, plus a real-time dashboard to monitor what your agent is doing.

npx claude-code-templates@latest

66 skills across 12 categories — languages, backend/frontend frameworks, infrastructure, APIs, testing, DevOps, security and data/ML — plus workflow commands for running development epics end to end.

/plugin marketplace add jeffallan/claude-skills
/plugin install fullstack-dev-skills@jeffallan

The canonical awesome-list for the Claude Code ecosystem — a huge, well-maintained index of skills, hooks, slash-commands, agent orchestrators and apps. The best place to discover what's out there.

A hand-picked catalog of Claude Code plugins so you can discover and install community skills from one place instead of hunting across GitHub.

/plugin marketplace add obra/superpowers-marketplace

The bleeding-edge testbed for Superpowers: newer skills and experimental techniques before they graduate into the main framework.

/plugin marketplace add obra/superpowers-lab

A large, organized collection of 100+ subagents covering frontend, backend, infra, data, security and more — drop the ones you need into your agents folder and delegate specialized work.

The actual agents a product studio uses day to day — engineering, design, product and ops roles — shared as a battle-tested starting point you can copy into your own setup.

A comprehensive, consistently formatted library of 100+ subagents for languages, frameworks and tools — a clean alternative collection to pull individual specialists from.

A pre-wired team of subagents — architect, frontend, backend, reviewer — that coordinate on a task, so you can hand off a feature instead of micromanaging one agent.

A tidy, curated set of custom subagents you can drop in — a good minimal starting point if the 100+ collections feel like too much.

Decompose a big task into pieces, hand them to coordinated agents and watch them work in a terminal visualization — multi-agent orchestration without building the plumbing yourself.

/plugin marketplace add affaan-m/claude-swarm

A hands-on reference for Claude Code's hook system — lifecycle events, safety gates and automation — with working examples you can lift to control exactly what your agent is allowed to do.

A shareable starter configuration with a structured CLAUDE.md "memory bank" — a proven layout for giving your agent durable project context from day one.

Lets the agent drive a real Chrome browser through the DevTools Protocol — navigate, click, read the DOM and debug live pages — useful for web testing and scraping tasks.

/plugin install superpowers-chrome@obra

A lightweight MCP server that gives the agent a private place to record reasoning and notes between turns — a simple, durable working memory you can plug into any MCP-capable tool.

npx -y private-journal-mcp

Gives agents recall of past sessions — what was tried, what worked, what failed — so they stop re-learning your codebase from scratch every time.

/plugin marketplace add obra/episodic-memory

A small plugin that evaluates, after each step, whether the agent has actually finished or should keep going — cutting the premature "all done!" that wastes a follow-up round.

/plugin install double-shot-latte@obra

Anthropic's reference MCP servers — filesystem, git, fetch, memory and more — the canonical building blocks for giving any agent safe, structured access to tools and data.

npx -y @modelcontextprotocol/server-filesystem ~/code

The go-to index of Model Context Protocol servers — databases, browsers, cloud APIs, search and hundreds more — to wire real-world capabilities into your agent.

Scans your agent setup for vulnerabilities — risky tool permissions, exposed MCP servers, dangerous configs — and ships as a CLI, a GitHub Action and a plugin so you can gate it in CI.

npx agentshield scan

Spin up background Claude Code sessions as workers in tmux and drive them programmatically — a simple way to fan work out across several sessions at once.

npx claude-session-driver