AI Agent 的
技能层。
让你的 AI Agent 自动发现、比较并安装正确的可复用 Skill。
Daily leaderboards
Skills moving now, ranked with evidence.
Daily snapshots combine capped activity signals, quality, trust, GitHub adoption, and real agent outcomes.
- 1design-creative
Frontend Design
7 views, 0 install copies, and 13 compares across 2 active days, with capped anti-spam weighting and 2 agent outcomes.
170K stars21 events / 2d - 2design-creative
Taste Skill: Anti-Slop Frontend
9 views, 1 install copies, and 15 compares across 2 active days, with capped anti-spam weighting and 2 agent outcomes.
78K stars29 events / 2d - 3agent-frameworks
Firecrawl
1 views, 0 install copies, and 12 compares across 2 active days, with capped anti-spam weighting and 2 agent outcomes.
139K stars13 events / 2d - 4data-analysis
D3
1 views, 0 install copies, and 33 compares across 2 active days, with capped anti-spam weighting and 1 agent outcomes.
113K stars34 events / 2d - 5agent-skills
Agent Skills
3 views, 0 install copies, and 25 compares across 2 active days, with capped anti-spam weighting and 1 agent outcomes.
88K stars28 events / 2d
Agent 解析
描述任务，得到一个安全的 Skill 方案。
Agent 执行前，API 会返回首选 Skill、备选项、策略判断、审计说明和安装计划。
任务匹配
96/100
推荐用于网页提取工作流
维护状态
活跃
Stars、更新、元数据和仓库健康度
安装审查
就绪
Agent 执行前的安全下一步
为什么是 OpenAgentSkill
别再把 Agent 扔进随机目录里。
Skill 注册表只有在 Agent 可以信任它时才真正有用。OpenAgentSkill 把分散的 GitHub 项目变成可排序、可审计、可安装的能力，供 Codex、Claude Code、Cursor、MCP 兼容 Agent 和自定义运行时调用。
- 01
从任务解析到 Skill
Agent 从意图出发，而不是从分类页出发。注册表把任务映射成首选 Skill、备选项和匹配理由。
- 02
安装前先看安全
Stars、更新、质量分、权限提示、风险和就绪说明，会和 Agent 将执行的命令一起返回。
- 03
人类浏览，Agent 调用 API
人可以浏览索引；Agent 可以通过解析、推荐和 Skill 端点调用同一套注册表。
注册表响应
一次调用，得到排序后的安装路径。
{
"task": "analyze stock news",
"agent_decision": {
"recommended_skill": "Last30days Skill",
"install_command": "npx skills add ...",
"why_recommended": [
"matches research workflow",
"strong Trust Score",
"audit warnings included"
],
"risk_summary": {
"safety": "review before install",
"notes": ["network access", "verify sources"]
}
}
}架构
从意图到安装之间的四层。
OpenAgentSkill 不是另一个静态列表，而是 Agent 在写文件、打开浏览器或安装第三方代码前可以调用的注册表循环。
Indexed
21,471
Signals
Fit · Risk
Surface
API · UI
- 01
Intent capture
A human or upstream agent describes the job in natural language.
Task · Agent · ContextIntent - 02
Recommendation engine
rankerSkills are ranked by workflow fit, maintenance, stars, and audit signals.
Fit · Quality · FreshnessRank - 03
Skill trust profile
Each candidate gets readiness notes, install commands, and review prompts.
Risk · Install · EvidenceAudit - 04
Agent install path
The registry returns the next action an agent can safely execute.
Codex · Claude Code · CursorInstall
Quickstart
From task description to install command.
- 01
Ask for a skill path
Resolve the task into one selected skill, alternatives, safety score, and install plan.
POST /api/agent/resolve - 02
Inspect the trust profile
Review fit, repository health, risks, and install readiness.
GET /api/agent/skills/crawl4ai - 03
Install in an agent workflow
Copy the command or hand it to Codex, Claude Code, Cursor, or a custom agent.
GET /api/skills/crawl4ai/install?format=text - 04
Automate discovery
Use the API as the registry layer behind your own agent runtime.
curl "https://www.openagentskill.com/api/agent/resolve?task=review+pull+requests&agent=codex"
对比
OpenAgentSkill 与其他 Skill 平台的区别。
核心判断很简单：普通目录是给人浏览的。OpenAgentSkill 是为了让 AI Agent 自动发现、比较并安装正确 Skill 而构建。
| Feature | OpenAgentSkill | skills.sh | agentskills.io | Native docs |
|---|---|---|---|---|
| Primary job | Recommend, compare, and install skills from one registry | Browse and install reusable agent skills | Define the open skill format and learning path | Explain skills inside each native agent platform |
| Agent-facing API | Yes - task-to-skill recommendations for agents | Directory and install workflow | Spec and documentation first | Platform-specific APIs and docs |
| Cross-agent positioning | Codex, Claude Code, Cursor, MCP-compatible agents, and custom tools | Open agent skills ecosystem | Open format for extending agents | Best for the vendor platform |
| Trust and audit signals | Stars, quality score, readiness notes, install review | Directory metadata | Metadata guidance in SKILL.md | Native platform controls |
| Best for | Letting an agent find the right skill automatically | Finding installable skills quickly | Learning or authoring the standard | Using skills in one product |
Comparison is based on each project's public positioning and documentation. The point is not that one project replaces another; OpenAgentSkill focuses on the registry and recommendation layer agents can call.
Workflow starts
Start from the job your agent needs to do.
Web scraping
Monitor pricing and extract tables
Coding agents
Inspect repos, patch bugs, verify changes
RAG workflows
Turn documents into grounded answers
Workflow automation
Connect repeated ops across tools
Skill layer
Registry for humans. Skill layer for agents.
Browse when you are exploring. Call the recommendation API when your agent needs to pick, compare, and install a skill automatically.