openclaw-rpa

分类: 数据与AI | 上传者: laziobird | 下载: 0 | 版本: v1.0（最新）

记录一次浏览器和本地文件操作；重播无需法学硕士即可运行——与人工智能浏览相比，节省成本，速度更快，没有幻觉。 github.com/laziobird/openclaw-rpa

更新日志: Source: GitHub https://github.com/laziobird/openclaw-rpa

目录结构

当前层级: tree/main/

📁 examples/
- 📄 README.md 756 B
📁 proofs/
- 📄 .gitignore 59 B
📁 rpa/
- 📄 onlineshoppingv1.py 6.1 KB
- 📄 onlineshoppingv1_playwright_commands.jsonl 7.6 KB
- 📄 wiki.py 5.4 KB
- 📄 wiki_playwright_commands.jsonl 5.4 KB
- 📄 yahoonew.py 6.0 KB
- 📄 yahoonew_playwright_commands.jsonl 9.7 KB
- 📄 每日美股快报整理v11.py 2.6 KB
- 📄 电商网站购物v10.py 5.7 KB
- 📄 电商网站购物v10_playwright_commands.jsonl 7.5 KB
- 📄 获取豆瓣电影数据.py 8.3 KB
- 📄 获取豆瓣电影数据_playwright_commands.jsonl 10.3 KB
📁 scripts/
- 📄 bootstrap_config.py 659 B
- 📄 export-clawhub-bundle.sh 1.4 KB
- 📄 install.sh 815 B
- 📄 set_locale.py 1.3 KB
📄 .gitignore 668 B
📄 config.example.json 130 B
📄 envcheck.py 4.1 KB
📄 LICENSE.md 11.1 KB
📄 LOCALE.md 822 B
📄 NOTICE.md 179 B
📄 playwright-templates.md 8.0 KB
📄 README.md 7.4 KB
📄 README.zh-CN.md 7.8 KB
📄 recorder_server.py 27.4 KB
📄 registry.json 213 B
📄 requirements.txt 97 B
📄 rpa_manager.py 35.4 KB
📄 SKILL.en-US.md 27.9 KB
📄 SKILL.md 5.6 KB
📄 SKILL.zh-CN.md 29.5 KB
📄 synthesis-prompt.md 6.2 KB

SKILL.md

---
name: openclaw-rpa
description: Record browser & local-file actions once; replay runs without the LLM—save $ vs AI browsing, faster, no hallucinations. github.com/laziobird/openclaw-rpa
metadata:
  openclaw:
    emoji: "🤖"
    os: ["darwin", "linux"]
    localeConfig: "config.json"
    instructionFiles:
      zh-CN: "SKILL.zh-CN.md"
      en-US: "SKILL.en-US.md"
---

# openclaw-rpa — **Locale router (read this first)**

**GitHub:** **[https://github.com/laziobird/openclaw-rpa](https://github.com/laziobird/openclaw-rpa)** — source, README, install, sample scripts under `rpa/`.

**Example flows** (ideas; record once, replay many times—**follow each site’s terms and local law**): **e‑commerce login & shopping**; **Yahoo Finance** stock quotes / news headlines; movie sites **reviews & ratings** in one automated run.

## What this skill does

**openclaw-rpa** turns repeatable **web** and optional **local file** work into a **Playwright Python** script by **recording** what actually happens in a real browser (plus file steps when needed). **Replay** runs that script directly—**not** the model clicking every time—so runs are **deterministic**, **cheaper**, and **less error-prone** than ad-hoc “automate this now” prompts.

**Why this matters**

1. **Saves compute and money** — Having a **large model** drive the browser on **every** run can cost **roughly single-digit to tens of US dollars** per heavy session (tokens, tools, long context). After you **record once**, repeat work **does not call the model**—replay is **much faster** and **near-zero** LLM cost for those steps.
2. **Verify once, run the same way every time** — During recording you **confirm** the flow works; later, replay **executes the saved steps** deterministically. You avoid asking the AI to “do it again” on every run, which **hurts consistency** and **raises hallucination risk**.

## When to use

| You want to… | Send |
|----------------|------|
| **Start recording** a new flow | `#automation robot`, `#RPA`, `#rpa`, or mention **Playwright automation** |
| **See saved tasks** you can run | `#rpa-list` |
| **Run a saved task** (e.g. new chat) | `#rpa-run:{task name}` |
| **Run in this chat** | `run:{task name}` (`zh-CN`: `#运行:{任务名}`) |

## Quick start (after install)

```text
#rpa-list
#rpa-run:your-task-name
```

Full protocol, state machine, `record-step` JSON, **progressive probing**, and **selector strength** (composite CSS — container + tag / attributes / `:has()`; avoid bare `h3`) live in the locale file below.

## Output

Generated file is **ordinary Python** (`rpa/*.py`) — runs standalone with `python3`, editable, no OpenClaw dependency at replay time.

## Scope

**Browser** — clicks, fill, select, scroll, wait, screenshot, text extraction.  
**Files (optional)** — `extract_text` writes to disk; patch `rpa/*.py` for folder / file ops after recording.  
**Out of scope** — large ETL, databases, heavy OS automation.

## Recommended sites

**Good fits** — predictable structure, works well out of the box:

| Category | Examples |
|----------|---------|
| Finance / data | Yahoo Finance, investing.com |
| E-commerce | Sauce Demo (`saucedemo.com`), AliExpress, eBay |
| News / media | BBC News, Reuters, Hacker News, Reddit listing pages |
| Reference | Wikipedia, GitHub public repo / issues pages |

**Not recommended** — likely to break or require manual intervention:

| Situation | Why |
|-----------|-----|
| Highly dynamic SPAs (heavy client-side routing) | Selectors shift between renders; snapshots may miss content |
| CAPTCHA / bot-detection (reCAPTCHA, hCaptcha, Cloudflare) | Automation blocked; human verification required. May be supported in the future. |
| Login-gated flows without saved sessions | Credentials / 2FA must be handled manually before replay |

> **Tip:** on a new site, start with `goto` + `snapshot` to confirm the page structure is readable before building a full flow.

## Mandatory: load the correct instruction file

1. **Read** `config.json` in this skill directory. If it does not exist, read **`config.example.json`** (same shape; default `locale` is **`en-US`**).
2. Read the `"locale"` field. Allowed values: **`zh-CN`** and **`en-US`** (repository default in **`config.example.json`**: **`en-US`**).
3. **Immediately use the Read tool** to load the **full** skill body:
   - `zh-CN` → **`SKILL.zh-CN.md`**
   - `en-US` → **`SKILL.en-US.md`**

4. **Follow only that file** for state machine, triggers, `record-step` JSON, onboarding text, and user-facing replies.

5. **Reply to the user in the active locale’s language:**
   - `zh-CN` → Simplified Chinese for agent messages (user may still type English).
   - `en-US` → English for agent messages (user may still type Chinese).

## Changing language

- Copy `config.example.json` → `config.json` if needed (`python3 scripts/bootstrap_config.py`), then edit `"locale"`, **or**
- Run: `python3 scripts/set_locale.py en-US` / `python3 scripts/set_locale.py zh-CN` (creates `config.json` from the example when missing).

After a locale change, the agent should **re-read** the matching `SKILL.*.md` in a new turn or session. See **README.md** in this directory for the full workflow.

## ClawHub / discovery

- **SKILL.md** (this file): short router + **when to use** + **quick start** for listings like [ClawHub](https://clawhub.ai/).
- **SKILL.zh-CN.md** / **SKILL.en-US.md**: full **onboarding**, **recording**, **RUN/LIST**, and anti-timeout rules.

## Relative paths

When the loaded file references `playwright-templates.md`, `synthesis-prompt.md`, or `rpa_manager.py`, resolve paths **relative to this skill directory** (parent of `SKILL.md`).

登录后下载/点赞/收藏 ❤ 45 | ★ 0

openclaw-rpa

目录结构

SKILL.md

举报内容

提示