Crawlio Plugin

<p align="center"> <img src=".github/banner-1.svg" alt="Crawlio AI Skills" width="650"> </p> <p align="center"> <a href="https://github.com/Crawlio-app/crawlio-plugin/releases"><img src="https://img.shields.io/github/v/release/Crawlio-app/crawlio-plugin?color=28c840&label=version" alt="Version"></a> <a href="LICENSE"><img src="https://img.shields.io/github/license/Crawlio-app/crawlio-plugin?color=8b949e" alt="License"></a> <a href="https://agentskills.io"><img src="https://img.shields.io/badge/standard-Agent_Skills-9b4dff" alt="Agent Skills"></a> <a href="https://crawlio.app"><img src="https://img.shields.io/badge/platform-macOS_15+-white" alt="macOS 15+"></a> </p>

---

Crawl, reverse-engineer, and analyze any website from your AI agent.

10 skills. 1 agent. One MCP server behind all of it.

Install

claude plugin install github:Crawlio-app/crawlio-plugin

Or add the server directly to any MCP client:

{
  "mcpServers": {
    "crawlio": {
      "command": "npx",
      "args": ["-y", "@crawlio/mcp@latest"]
    }
  }
}

Works with Claude Code, Gemini CLI, Cursor, Windsurf. Anything that speaks MCP.

No build step. No binary. npx handles it.

Skills

Skill	What it does
crawlio-mcp	Entry point. Explains the server, routes you to the right skill
crawl-site	Crawl with config, monitoring, and retry
extract-and-export	Crawl, extract, export. 7 formats
observe	Query the observation timeline
finding	Create and query evidence-backed findings
audit-site	Multi-pass site audit with findings report
web-research	Structured acquire, normalize, analyze pipeline
decompile-spa	Reverse-engineer an SPA: modules, routes, state
extract-secrets	Scan JS bundles for hardcoded credentials. Authorized targets only
design-system	Extract design tokens from a live page. Colors, type, spacing, breakpoints

Skills are plain Markdown. The MCP server does the work.

Agent

Site Auditor — four passes. Recon, crawl, analyze, report. Evidence-backed findings with prioritized fixes.

See agents/site-auditor.md.

Architecture

Your AI agent
    │
    ▼
@crawlio/mcp  ─────►  5 abstract tools
(aggregator)           discover / call / do / cortex / consult
    │
    ├── chrome           browser extension, live context
    ├── headless         crawlio-agent-headless, RE skills
    ├── app              Crawlio.app, persistent projects
    └── headless-engine  fallback, zero dependencies

Skills reference concrete tool names. The aggregator routes them to the right pillar. Add a pillar, every skill gets the upgrade. No plugin release needed.

Crawlio.app adds persistent projects and the full UI. Without it, the aggregator falls back to headless-engine. Crawling still works. Projects live in memory instead of on disk.

The Chrome extension adds live browser context: framework detection, network interception, console logs. Without it, those skills fall back to headless Chromium.

Both optional. The server runs either way.

Structure

crawlio-plugin/
├── .claude-plugin/
│   └── plugin.json
├── .mcp.json
├── skills/
│   ├── crawlio-mcp/
│   ├── crawl-site/
│   ├── extract-and-export/
│   ├── observe/
│   ├── finding/
│   ├── audit-site/
│   ├── web-research/
│   ├── decompile-spa/
│   ├── extract-secrets/
│   └── design-system/
├── agents/
│   └── site-auditor.md
├── FORKING.md
└── LICENSE

Each skill is a single SKILL.md inside its directory. Each file follows the Agent Skills open standard.

Fork it

This plugin is designed to be copied.

SEO auditor. Security scanner. Competitive teardown. Content migration planner. Pick a domain, keep the tools, rewrite the skills.

See FORKING.md.

License

MIT