Skip to content

20 Essential Claude Code Repos — Ecosystem Survey

Summary

Curated survey of 20 Claude Code ecosystem repos across five categories: official Anthropic, awesome-lists, skills, subagents, MCP servers, and orchestration/memory. The standout for our toolkit work is obra/superpowers (148k★) which implements brainstorm → spec → plan → TDD → review → merge with built-in subagent orchestration — direct conceptual parallel to the /vt-c-activate → /vt-c-2-plan → /vt-c-3-build → /vt-c-4-review chain. Worth a comparative evaluation.

Key Details

  • Official Anthropic: anthropics/claude-code (55k★), anthropics/skills (37.5k★ — PDF/DOCX/XLSX/PPTX/image skills), anthropics/anthropic-cookbook, anthropics/claude-code-action (GitHub Action for @claude in PR comments)
  • Awesome-lists: hesreallyhim/awesome-claude-code (28.5k★, only Claude itself can submit PRs), ComposioHQ/awesome-claude-skills (50+ production skills), ComposioHQ/awesome-claude-plugins
  • Skills: obra/superpowers (148k★ — full workflow chain with subagents), travisvn/awesome-claude-skills (22k+ installs, warns about untrusted skills)
  • Subagents: wshobson/agents (25k★ — strategy/dev/security/design/data/research), baryhuang/claude-code-agents (59 specialized agents incl. TypeScript/Python/Java/Kotlin reviewers)
  • MCP servers: github/github-mcp-server (most-used), microsoft/playwright-mcp (browser automation), crystaldba/postgres-mcp (read-only by default), upstash/context7 (real-time library docs — we already use this), zilliztech/claude-context (semantic code search)
  • Memory/orchestration: ruvnet/claude-flow (11.4k★ — multi-agent + persistent memory), smtg-ai/claude-squad (5.6k★ — terminal multiplexer for parallel sessions), thedotmack/claude-mem (context compression for long-term memory)
  • TDD enforcement: nizos/tdd-guard (1.7k★ — blocks Claude from skipping tests)
  • Discovery hub: claudewave/claudewave (tracks 3,800+ repos, updates every 12h)

Why Rolf Thinks This Matters

Further Reading