cheliped-browser

分类: 工具与效率 | 上传者: tykimos | 下载: 0 | 版本: v1.0（最新）

Agent Browser Runtime for browsing, observing, and interacting with web pages via Chrome DevTools Protocol (CDP). Use this skill when Claude needs to: (1) navigate to and browse websites, (2) extract page content, text, or links from any URL, (3) fill forms or type into input fields, (4) click buttons or links on a page, (5) take screenshots of web pages, (6) execute JavaScript in a page context, (7) perform login or search actions on websites, (8) scrape or crawl web content, or any other browser-based task. Triggers on: browse, crawl, scrape, web page, website, navigate, open URL, screenshot a site, fill a form online, login to a site.

更新日志: Source: GitHub https://github.com/tykimos/cheliped-browser

目录结构

当前层级: 根目录

📁 docs/
- 📁 images/
  - 📄 benchmark-openclaw-speed.png 63.9 KB
  - 📄 benchmark-openclaw-tokens.png 81.8 KB
  - 📄 benchmark-quality-breakdown.png 58.4 KB
  - 📄 benchmark-speed-per-site.png 51.6 KB
  - 📄 benchmark-summary.png 83.9 KB
  - 📄 benchmark-tokens-per-site.png 59.7 KB
  - 📄 config-maxlinks.png 86.4 KB
  - 📄 config-maxtextlength.png 83.7 KB
  - 📄 config-recall-persite.png 70.7 KB
📁 scripts/
- 📁 examples/
  - 📄 advanced-usage.ts 2.1 KB
  - 📄 basic-usage.ts 1.1 KB
  - 📄 browse-demo.ts 2.4 KB
  - 📄 instagram-demo.ts 1.6 KB
  - 📄 instagram-login-demo.ts 3.9 KB
- 📁 src/
  - 📁 api/
    
    📄 cheliped.ts 11.5 KB
    
    📄 index.ts 42 B
  - 📁 browser/
    
    📄 controller.ts 13.8 KB
    
    📄 index.ts 87 B
    
    📄 page.ts 3.8 KB
  - 📁 cdp/
    
    📄 connection.ts 5.5 KB
    
    📄 index.ts 187 B
    
    📄 launcher.ts 5.9 KB
    
    📄 transport.ts 4.0 KB
    
    📄 typed-cdp.ts 4.5 KB
  - 📁 dom/
    
    📄 agent-dom.ts 4.0 KB
    
    📄 compressor.ts 4.0 KB
    
    📄 extractor.ts 6.3 KB
    
    📄 filter.ts 2.9 KB
    
    📄 index.ts 256 B
    
    📄 semantic.ts 6.0 KB
  - 📁 graph/
    
    📄 action-generator.ts 6.8 KB
    
    📄 action.types.ts 552 B
    
    📄 index.ts 336 B
    
    📄 ui-graph.ts 6.8 KB
    
    📄 ui-graph.types.ts 818 B
  - 📁 security/
    
    📄 exfiltration-guard.ts 1.8 KB
    
    📄 index.ts 256 B
    
    📄 prompt-guard.ts 951 B
    
    📄 security-layer.ts 3.2 KB
    
    📄 security.types.ts 645 B
  - 📁 session/
    
    📄 cookie-store.ts 987 B
    
    📄 index.ts 191 B
    
    📄 session-manager.ts 2.5 KB
    
    📄 session.types.ts 472 B
  - 📁 types/
    
    📄 agent-dom.types.ts 724 B
    
    📄 api.types.ts 688 B
    
    📄 index.ts 331 B
    
    📄 internal-dom.types.ts 1.2 KB
    
    📄 options.types.ts 1.1 KB
  - 📄 index.ts 1.4 KB
- 📁 tests/
  - 📁 fixtures/
    
    📄 expected-agent-dom.json 871 B
    
    📄 sample-dom.html 1.1 KB
  - 📁 helpers/
    
    📄 html-to-internal.ts 2.4 KB
  - 📁 integration/
    
    📄 cdp-connection.test.ts 1.1 KB
    
    📄 dom-extraction.test.ts 2.2 KB
    
    📄 navigation.test.ts 1.2 KB
    
    📄 security.test.ts 1.1 KB
    
    📄 ui-graph.test.ts 1.2 KB
  - 📁 unit/
    
    📁 cdp/
    
    📄 transport.test.ts 5.9 KB
    
    📄 typed-cdp.test.ts 3.0 KB
    
    📁 dom/
    
    📄 agent-dom.test.ts 5.9 KB
    
    📄 compressor.test.ts 5.9 KB
    
    📄 filter.test.ts 8.0 KB
    
    📄 semantic.test.ts 7.6 KB
    
    📁 graph/
    
    📄 action-generator.test.ts 4.6 KB
    
    📄 ui-graph.test.ts 3.9 KB
    
    📁 security/
    
    📄 prompt-guard.test.ts 1.8 KB
    
    📄 security-layer.test.ts 3.1 KB
- 📄 .npmrc 29 B
- 📄 .prettierrc 106 B
- 📄 benchmark-challenge.mjs 19.8 KB
- 📄 benchmark-compare.mjs 15.2 KB
- 📄 benchmark-limitations.mjs 19.9 KB
- 📄 benchmark-quality.mjs 26.5 KB
- 📄 benchmark.mjs 8.4 KB
- 📄 cheliped-cli.mjs 9.1 KB
- 📄 generate-charts.py 7.2 KB
- 📄 package.json 1.1 KB
- 📄 pnpm-lock.yaml 77.5 KB
- 📄 pnpm-workspace.yaml 35 B
- 📄 tsconfig.json 516 B
- 📄 tsup.config.ts 192 B
- 📄 vitest.config.ts 157 B
- 📄 vitest.integration.config.ts 264 B
📄 .gitignore 64 B
📄 LICENSE.txt 1.0 KB
📄 README.md 42.3 KB
📄 SKILL.md 3.3 KB

SKILL.md

---
name: cheliped-browser
description: "Agent Browser Runtime for browsing, observing, and interacting with web pages via Chrome DevTools Protocol (CDP). Use this skill when Claude needs to: (1) navigate to and browse websites, (2) extract page content, text, or links from any URL, (3) fill forms or type into input fields, (4) click buttons or links on a page, (5) take screenshots of web pages, (6) execute JavaScript in a page context, (7) perform login or search actions on websites, (8) scrape or crawl web content, or any other browser-based task. Triggers on: browse, crawl, scrape, web page, website, navigate, open URL, screenshot a site, fill a form online, login to a site."
---

# Agent Browser Runtime

Control Chrome via CDP. Extract **Agent DOM** — a compressed DOM where every interactive element gets a numeric `agentId`.

## Setup

Run once before first use:

```bash
cd scripts && npm install && npm run build
```

## Core Workflow: Observe-Act Loop

```bash
# 1. Navigate and observe
node scripts/cheliped-cli.mjs '[{"cmd":"goto","args":["https://example.com"]},{"cmd":"observe"}]'

# 2. Use agentId from observe output to interact
node scripts/cheliped-cli.mjs '[{"cmd":"fill","args":["3","search term"]},{"cmd":"click","args":["4"]},{"cmd":"observe"}]'
```

First call auto-launches Chrome and saves session. Subsequent calls reconnect. Call `close` when done.

## Commands

| Command | Args | Returns |
|---------|------|---------|
| `goto` | `["url"]` | `{ success, url, title }` |
| `observe` | none | `{ nodes, texts, links }` — `nodes[].id` = agentId |
| `click` | `["agentId"]` | `{ success, action, agentId }` |
| `fill` | `["agentId", "text"]` | `{ success, action, agentId }` |
| `screenshot` | `["path"]` (optional) | `{ success, path, size }` |
| `run-js` | `["expression"]` | `{ success, result }` |
| `extract` | `["text"\|"links"\|"all"]` | `{ type, data }` |
| `actions` | none | `[{ id, type, label, params }]` |
| `perform` | `["actionId"]` + `"params":{...}` | `{ success, actionId }` |
| `observe-graph` | none | `{ nodes, edges, forms }` |
| `close` | none | `{ success }` |

## Examples

### Browse and extract content

```bash
node scripts/cheliped-cli.mjs '[{"cmd":"goto","args":["https://news.ycombinator.com"]},{"cmd":"observe"}]'
```

### Fill a form and submit

```bash
node scripts/cheliped-cli.mjs '[{"cmd":"observe"},{"cmd":"fill","args":["7","hello"]},{"cmd":"click","args":["8"]}]'
```

### Semantic action (login, search)

```bash
node scripts/cheliped-cli.mjs '[{"cmd":"goto","args":["https://example.com/login"]},{"cmd":"actions"}]'
node scripts/cheliped-cli.mjs '[{"cmd":"perform","args":["login-form"],"params":{"email":"user@example.com","password":"pass"}}]'
```

### Concurrent sessions

```bash
node scripts/cheliped-cli.mjs --session agent1 '[{"cmd":"goto","args":["https://site-a.com"]}]'
node scripts/cheliped-cli.mjs --session agent2 '[{"cmd":"goto","args":["https://site-b.com"]}]'
```

## Key Notes

- Call `observe` before `click`/`fill` — agentIds are only valid after observation.
- `fill` works with React/SPA apps (uses native input value setters).
- Agent DOM is token-compressed — far fewer tokens than raw HTML.
- Chrome persists between calls until `close`. No restart needed.
- All output is JSON to stdout. Errors: `{ "error": "message" }`.

登录后下载/点赞/收藏 ❤ 4 | ★ 0

cheliped-browser

目录结构

SKILL.md

举报内容

提示