sickn33 / agentic-awesome-skills

A local, agent-first control plane where AI coding agents own their skill selection from a catalog of 1,987+ agentic skills.

活跃维护 MIT Python Tracked
45.7k 6.7k 1 天前
CINPM
agentic-skills ai-agents antigravity claude-code mcp ai-workflows codex-cli developer-tools gemini-cli skill-library agent-skills ai-agent-skills antigravity-skills claude-code-skills codex-skills cursor-skills gemini-skills ai-coding cursor kiro

星标趋势

数据积累中,暂无足够数据生成趋势图

AI 分析

项目摘要

AAS Core is a local, agent-first control plane that provides a catalog of 1,987+ agentic skills for AI coding agents like Claude Code, Codex CLI, Gemini CLI, and Cursor. It enables agent-owned skill selection, stack validation, and immutable planning through a local MCP server and CLI, with the AI agent (not the system) choosing which skills to apply.

为什么值得关注

With 44K+ stars and 301 contributors, AAS Core has rapidly become a major ecosystem for AI coding agent skills. Its 'agent-owned' selection paradigm—where the AI agent inspects the project and chooses exact skills rather than being recommended them—represents a novel approach to agent-tool integration, backed by a local-first architecture with MCP support.

优势

  • Massive community adoption with 44K+ stars and 301 contributors
  • Multi-agent compatibility (Claude Code, Codex CLI, Gemini CLI, Cursor, Kiro)
  • Local-first, read-only architecture with MCP integration
  • Frequent releases (100 in 6 months) indicating active development
  • Agent-owned selection paradigm gives control to the AI agent

局限性

  • No automated tests despite large scale
  • No Docker support or official examples
  • Apply and recovery features remain experimental
  • 0 open issues is unusual and may indicate issue tracking problems

使用场景

  • Discovering and managing skills for AI coding agents
  • Validating agent-selected skill stacks before application
  • Creating reproducible, reviewable agent development plans
  • Building standardized agent configurations across teams
目标用户: Developers and teams using AI coding assistants (Claude Code, Codex, Gemini CLI, Cursor) who want a structured, local-first way to manage and validate agent skills
学习曲线:
分析模型:LongCat-2.0 | 分析时间:1 个月前