2.6k
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill clawme- _meta.json273 B
- SKILL.md3.4 KB
Overview
This skill sends actionable instructions to a user's real Chrome browser through the ClawMe extension so the user can review and confirm each action. It operates in the user’s actual browser session (with their logged-in accounts), enabling form fills, clicks, content extraction, tweets, emails, and multi-step workflows. The user always sees each instruction in a side panel and taps Execute to run it.
How this skill works
The agent issues instructions via an HTTP POST or the provided clawme_send tool to the ClawMe backend, including a token that matches the installed Chrome extension. Instructions specify a type (compose_tweet, fill_form, click, extract, etc.) and a payload with selectors, URLs, or content. Each instruction appears in the extension side panel; the user confirms execution, and extracted results are returned to the agent.
When to use it
- Automate repetitive browser tasks that require the user’s authenticated session
- Fill complex web forms using precise CSS selectors or contenteditable fields
- Trigger multi-step workflows where the user wants visibility and control at each step
- Extract specific page content and return it to the agent for processing
- Open links or initiate actions (tweet, email) from the user’s real account
Best practices
- Provide clear, minimal instructions and use precise CSS selectors to target fields or buttons
- Include url in payload when action should run on a specific page; omit url to target the current tab
- Use workflow metadata (workflow_id and step) for multi-step sequences so the extension shows progress
- Test selectors and flows locally (CLAWME_BASE_URL can point to a local instance) before large batches
- Avoid sending sensitive tokens in instruction payloads; use the configured extension token via X-ClawMe-Token
Example use cases
- Pre-fill an onboarding form across multiple pages using chained fill_form and click steps
- Compose and post a tweet using the user’s logged-in X account via compose_tweet
- Extract product listings or search results from a page and return text for downstream summarization
- Open a target URL, fill a sign-up form, and submit it as a guided workflow the user can run step-by-step
- Draft and send an email in Gmail using compose_email with use_gmail:true
FAQ
The user installs the ClawMe Chrome extension and configures the Backend URL and a client token; the agent uses that token in X-ClawMe-Token headers.
How are multi-step workflows handled?
Include meta.workflow_id and meta.step in each instruction. The extension shows a workflow card with a progress bar and allows sequential execution.