Pithelo

Five agent repos gaining momentum. Every one measured.

Ranked by stars gained over the last 7 days, growth rate and recent commit activity.

Rank 1

alirezarezvani/claude-skills

An open-source library of roughly 398 reusable skill packages for AI coding agents, each built from a SKILL.md instruction file, stdlib-only Python CLI scripts, and reference docs. It covers engineering, marketing, product, compliance, C-level advisory, and research, and installs on Claude Code, Codex, Gemini CLI, Cursor, and about nine other coding tools.

Developers who want ready-made, cross-agent skill, agent, and slash-command packages to adopt or use as templates for building their own should look at it.

  • ★ 28,927
  • +★ 415 in 7 days
  • 1.5% growth
  • MIT
  • README ≈ 10,973 tokens to load into context

AI-written summary

Score91.0
Rank 2

wanshuiyin/Auto-claude-code-research-in-sleep

ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works with Claude Code, Codex, OpenClaw, or any LLM agent.

  • ★ 18,250
  • +★ 389 in 7 days
  • 2.2% growth
  • MIT

AI-written summary

Score87.0
Rank 3

xpzouying/xiaohongshu-mcp

xiaohongshu-mcp is an MCP server that lets AI assistants interact with xiaohongshu.com (小红书). Its documented tools cover login and status checks, posting image-text and video notes, searching, fetching the home feed, reading post details and comments, posting and replying to comments, viewing user profiles, and liking or favoriting notes.

Developers building agents that need to read or operate a Xiaohongshu account should look at it, noting that it requires login, that one account cannot be logged in on multiple web clients at once, and that the README suggests the x-mcp browser extension if Docker deployment proves difficult.

  • ★ 16,637
  • +★ 207 in 7 days
  • 1.3% growth
  • Apache-2.0
  • README ≈ 21,527 tokens to load into context

AI-written summary

Score66.0
Rank 4

triggerdotdev/trigger.dev

Trigger.dev is an open-source TypeScript platform for building and deploying durable AI agents and workflows as long-running tasks. It provides retries, queues, checkpointing, waitpoints for human-in-the-loop approval, realtime run streaming, observability, and elastic scaling without execution timeouts.

Developers building agents or background jobs that run longer than serverless timeouts allow, or that need retries, human approval steps, or custom runtimes like FFmpeg, Python, or browsers, should look at it.

  • ★ 16,647
  • +★ 52 in 7 days
  • 0.3% growth
  • Apache-2.0
  • README ≈ 3,638 tokens to load into context

AI-written summary

Score64.0
Rank 5

modelcontextprotocol/python-sdk

The official Python SDK for the Model Context Protocol (MCP), requiring Python 3.11+. It lets you build MCP servers that expose tools, resources, and prompts, and MCP clients that connect to any MCP server, over stdio, Streamable HTTP, and SSE transports.

Developers who want to expose functions or data to LLM applications as MCP servers, or connect an agent to existing MCP servers, can use it to do so from type-hinted Python functions without writing JSON Schema or protocol handling.

  • ★ 24,701
  • +★ 65 in 7 days
  • 0.3% growth
  • MIT
  • README ≈ 1,936 tokens to load into context

AI-written summary

Score61.0