screenshot-workflow

通过浏览器工具打开网址、截取屏幕截图，并借助消息工具将截图发回的确定性工作流。

SKILL.md

--- frontmatter

name: screenshot-workflow
description: Deterministic workflow for opening a URL, taking screenshots via the browser tool, and sending them back via the message tool.
user-invocable: true
metadata: {"openclaw":{"emoji":"📸","requires":{"config":["browser.enabled"]}}}

Screenshot Workflow (Deterministic)

Use this workflow whenever the user requests a screenshot ("screenshot", "take a screenshot", "截图", "截个图") and provides a URL.

Inputs

•
A URL that may be:
- •full (https://...)
- •starts with www
- •bare domain (example.com)

Workflow (must follow)

Step 1 — Normalize URL

•If the URL does not start with http/https, prepend https://.
•Call the normalized URL TARGET_URL.

Step 2 — Open page

•Use the browser tool to open TARGET_URL.
•If redirected, treat the final landing URL as FINAL_URL.

Step 3 — Screenshot

•Use the browser tool to capture a screenshot (prefer full-page).
•
If full-page capture fails or is unsupported:
- •capture viewport screenshot
- •optionally scroll and capture up to 3 total images

Step 4 — Send back to user

•
Use the message tool to send:
- •Text: Screenshot captured: <FINAL_URL>
- •Attach 1–3 screenshot images

Required failure handling

•
If open/connect fails:
- •retry once (optionally check browser status first)
- •if still failing, report the concrete reason (CDP unreachable, browser not configured, attach required, permissions/sandbox)
•
If the site requires login/CAPTCHA/human verification:
- •stop automation
- •inform the user manual interaction is required
- •do NOT fabricate results

Tool-call budget

•Max 8 tool calls total.