← Back to Skills Marketplace
dongsheng123132

ClawMe

by dongsheng123132 · GitHub ↗ · v1.0.0
cross-platform ✓ Security Clean
332
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install clawme
Description
Send instructions to user's real browser via ClawMe Chrome extension. Fill forms, tweet, email, click, extract — user sees and confirms each action in the si...
README (SKILL.md)

ClawMe — Execute Actions in User's Real Browser

Send instructions to the user's Chrome browser via the ClawMe extension. The user sees each instruction in a side panel and clicks "Execute" to confirm. Unlike headless browsers, ClawMe runs in the user's real browser with their actual login sessions.

Setup: User installs ClawMe Chrome extension, configures Backend URL and Token.

How to Send Instructions

Use the clawme_send tool if available. Otherwise, make an HTTP POST:

POST ${CLAWME_BASE_URL:-https://api.clawme.net}/v1/instructions
Headers: Content-Type: application/json, X-ClawMe-Token: ${CLAWME_CLIENT_TOKEN}
Body: {"target":"browser","instruction":{"type":"\x3Ctype>","payload":{...}}}

Instruction Types

compose_tweet

When user says "tweet about...", "post on X..."

{"type":"compose_tweet","payload":{"text":"tweet content"}}

compose_email

When user says "write email to...", "send email..."

{"type":"compose_email","payload":{"to":"[email protected]","subject":"Subject","body":"Body text","use_gmail":true}}

fill_form

When user says "fill the form...", "enter my info..."

{"type":"fill_form","payload":{"url":"https://example.com/form","fields":{"#name":"John","input[name=email]":"[email protected]","select[name=country]":"US"}}}

Supports: inputs, textareas, selects, checkboxes, radio, contenteditable (Xiaohongshu, Medium). Use CSS selectors as field keys. Omit url for current page.

click

When user says "click the button...", "submit the form..."

{"type":"click","payload":{"selector":"button[type=submit]","url":"https://example.com/form"}}

extract

When user says "get the text from...", "scrape..."

{"type":"extract","payload":{"selector":".results","url":"https://example.com/search"}}

Result (extracted text) is reported back to the agent.

open_url

{"type":"open_url","payload":{"url":"https://example.com","in_new_tab":true}}

remind

{"type":"remind","payload":{"title":"Meeting","body":"Team standup in 5 minutes"}}

Multi-Step Workflows

Chain instructions with meta.workflow_id and meta.step:

POST /v1/instructions — {"target":"browser","instruction":{"type":"open_url","payload":{"url":"..."}}, "meta":{"workflow_id":"signup","step":1}}
POST /v1/instructions — {"target":"browser","instruction":{"type":"fill_form","payload":{"fields":{...}}}, "meta":{"workflow_id":"signup","step":2}}
POST /v1/instructions — {"target":"browser","instruction":{"type":"click","payload":{"selector":"button[type=submit]"}}, "meta":{"workflow_id":"signup","step":3}}

User sees a workflow card with progress bar and can execute all steps sequentially.

Environment Variables

  • CLAWME_CLIENT_TOKEN (required) — matches the token configured in the Chrome extension
  • CLAWME_BASE_URL (optional) — default https://api.clawme.net, or http://127.0.0.1:31871 for local
Usage Guidance
This skill is internally consistent with its description, but it grants an agent the ability to control your real browser and read page content through the ClawMe extension. Before installing: (1) Only install the official, reviewed ClawMe extension from a trusted source; (2) Treat CLAWME_CLIENT_TOKEN as highly sensitive — do not expose it to untrusted agents or public repos; (3) Confirm how the extension enforces per-action user confirmation and whether workflows can be bulk-approved; (4) Prefer manual approval for each action that accesses sensitive pages (banking, email, developer consoles); (5) If you need stronger guarantees, avoid storing the token in shared environment variables or restrict agent autonomy when this skill is enabled.
Capability Analysis
Type: OpenClaw Skill Name: clawme Version: 1.0.0 The skill bundle defines a browser automation tool that allows an AI agent to interact with the user's Chrome browser (e.g., fill forms, compose emails, extract text). A critical security feature is explicitly stated: 'user sees and confirms each action in the side panel.' This means no action can be performed without the user's explicit consent, significantly mitigating risks of unauthorized data exfiltration or malicious execution. The `SKILL.md` provides clear instructions for using the tool's API and does not contain any prompt injection attempts, obfuscated payloads, or instructions for the agent to perform harmful or stealthy actions. The `extract` capability, while powerful, is also subject to user confirmation and reports data back to the agent, not an arbitrary external endpoint defined by the skill itself.
Capability Assessment
Purpose & Capability
Name/description (remote control of a Chrome extension to fill forms, click, extract, post) matches the declared requirement (CLAWME_CLIENT_TOKEN) and the SKILL.md instructions (POST to the extension backend with that token). No unrelated env vars, binaries, or install steps are requested.
Instruction Scope
The runtime instructions let the agent request actions that run in the user's real browser (fill forms, click, compose email/tweet, extract page content). The SKILL.md says the user 'sees and confirms' each action, but it's ambiguous whether multi-step workflows can be bulk-approved. Because 'extract' returns page content to the agent and actions run under the user's live sessions, this capability can exfiltrate sensitive data or perform actions on behalf of the user if the token or confirmation flow is misused.
Install Mechanism
Instruction-only skill (no install spec, no downloaded code). This minimizes installation-side risk; nothing is written to disk by the skill bundle itself.
Credentials
Only CLAWME_CLIENT_TOKEN (and optional CLAWME_BASE_URL) are required, which is proportionate to controlling an extension. However, that single token effectively authorizes sending commands to the user's extension and should be treated as highly sensitive (it grants browser-action capability and potential access to page contents and logged-in sessions).
Persistence & Privilege
The skill does not request 'always: true' or any system-wide persistence. It is user-invocable and can be invoked by the agent (normal platform default).
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install clawme
  3. After installation, invoke the skill by name or use /clawme
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of ClawMe: Control your real Chrome browser via simple API instructions. - Send browser automation commands (fill forms, tweet, email, click, extract data) through the ClawMe Chrome extension. - Users see each instruction and confirm before execution, preserving control and security. - Supports multi-step workflows with progress tracking in the browser side panel. - Compatible with user's existing Chrome sessions and logins. - Simple setup: install extension and configure a client token.
Metadata
Slug clawme
Version 1.0.0
License
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is ClawMe?

Send instructions to user's real browser via ClawMe Chrome extension. Fill forms, tweet, email, click, extract — user sees and confirms each action in the si... It is an AI Agent Skill for Claude Code / OpenClaw, with 332 downloads so far.

How do I install ClawMe?

Run "/install clawme" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is ClawMe free?

Yes, ClawMe is completely free (open-source). You can download, install and use it at no cost.

Which platforms does ClawMe support?

ClawMe is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created ClawMe?

It is built and maintained by dongsheng123132 (@dongsheng123132); the current version is v1.0.0.

💬 Comments