Great AI & Skill repos
A curated list of GitHub repositories worth using: Agent Skills, MCP servers, coding agents, AI automation and LLMOps — one line each, plus our write-up where we have one.
37 repos · updated Oct 28, 2026
Agent Skills & Claude Code plugins
Skills, subagents and plugins for Claude Code (many also work with Codex, Cursor).
-
anthropics/skills
Anthropic's official skills repo: documents (docx, pdf, xlsx), design, skill-creator. The place to learn how SKILL.md is written.
Read post -
obra/superpowers
A skill set that makes the agent follow a process: brainstorm, plan, write tests first, then code.
Read post -
JuliusBrussee/caveman
A skill that makes Claude answer tersely to cut output tokens.
Read post -
cloudflare/security-audit-skill
Cloudflare's multi-phase security audit skill; findings are independently verified and machine-readable.
Read post -
wshobson/agents
A large plugin marketplace: dozens of agents, skills and commands by domain (backend, DevOps, security…). Pick selectively.
-
VoltAgent/awesome-claude-code-subagents
100+ ready-made specialised subagents; useful to study prompt and tool scoping for subagents.
MCP servers
Connect agents to browsers, GitHub, docs and outside tools.
-
modelcontextprotocol/servers
Reference servers from the MCP project (filesystem, git, fetch, memory…). Read them to see how an MCP server is built.
Read post -
github/github-mcp-server
GitHub's official MCP: issues, PRs, Actions; enable by toolset to save context.
Read post -
microsoft/playwright-mcp
Lets agents drive a browser through accessibility snapshots instead of screenshots.
Read post -
upstash/context7
Feeds version-correct library docs into the prompt so the AI stops using outdated APIs.
Read post -
oraios/serena
Symbol-level code reading and editing (IDE-style) for coding agents, instead of reading whole files.
Read post -
ahujasid/blender-mcp
Drive Blender with natural language; a good example of MCP for desktop software.
Read post -
punkpeye/awesome-mcp-servers
Community catalogue of MCP servers by domain. Review the code before installing unknown servers.
Coding agents & spec-driven
Coding agents, SDKs, and tools that keep agents on spec and on budget.
-
anthropics/claude-code
The Claude Code repo: issues, changelog, plugin and hook examples.
Read post -
anthropics/claude-code-action
GitHub Action: mention @claude in issues/PRs to review, fix bugs, answer questions.
Read post -
anthropics/claude-agent-sdk-python
Python SDK for building your own agents on the same engine as Claude Code.
Read post -
Fission-AI/OpenSpec
Lightweight spec-driven development: agree the change in a spec before the AI codes.
Read post -
github/spec-kit
GitHub's spec-driven toolkit: specify → plan → tasks → implement, works with many coding agents.
-
eyaltoledano/claude-task-master
Turns a PRD into dependency-aware tasks for an agent to work through.
-
Aider-AI/aider
AI pair programming in the terminal with automatic git commits; works with many models.
-
rtk-ai/rtk
Compresses terminal command output before it reaches the agent to save tokens.
Read post -
Graphify-Labs/graphify
Turns a codebase into a knowledge graph so the agent doesn't re-read it every session.
Read post
AI automation & web data
Automated workflows, browser agents, web data for LLMs.
-
n8n-io/n8n
Self-hostable workflow platform with an AI Agent node; the base for most automation posts on binhAI.
Read post -
browser-use/browser-use
An agent that drives the browser toward a goal written in plain language.
Read post -
unclecode/crawl4ai
Open-source crawler that turns web pages into clean Markdown for LLMs.
-
firecrawl/firecrawl
Web scrape/crawl API for agents; cloud and self-hosted options.
-
openclaw/openclaw
Self-hosted AI assistant you message via Telegram/WhatsApp. Think hard about its access first.
Read post
LLMOps: gateway, eval, observability
Running AI apps in production: cost control, prompt tests, tracing.
-
BerriAI/litellm
One API for many providers, with virtual keys and per-team budgets.
Read post -
promptfoo/promptfoo
Write tests for prompts, run evals and red-teaming in CI.
Read post -
langfuse/langfuse
Open-source tracing, evals and prompt management for LLM apps; self-hostable.
Learning & curated lists
Learning material, cookbooks and awesome lists to explore further.
-
anthropics/claude-cookbooks
Official example notebooks: tool use, RAG, prompt caching, agents.
-
luongnv89/claude-howto
Step-by-step Claude Code learning path, with a Vietnamese edition.
Read post -
hesreallyhim/awesome-claude-code
Curated Claude Code resources: commands, workflows, hooks, tools.
-
travisvn/awesome-claude-skills
Community skills and resources about Claude Skills.
-
VoltAgent/awesome-design-md
DESIGN.md samples so agents build UI that follows a design system.
Read post
Video & creative
Let agents build video and motion graphics in code.
-
heygen-com/hyperframes
Write video in HTML, render to MP4; built for agents.
Read post -
remotion-dev/skills
Remotion's official skills so Claude Code builds React videos correctly.
Read post
Never miss a new article
Follow binhAI via RSS (Feedly, Inoreader, Slack…) to get new articles as soon as they’re published.