use-web-skill
When working on a live webpage with browser tooling, check the document <head> before deeper interaction.
Workflow
- Read the page head and look for:
<link rel="web-skill" ...>- especially
hrefandtitle
- Compare each link
titleagainst the current task. - If a title looks relevant, fetch the linked
SKILL.md. - Read the downloaded skill and apply its instructions for the rest of the browser task.
- If no relevant
web-skilllink exists, continue with the normal browser workflow.
Selection rules
- Prefer the most task-specific title.
- Prefer exact domain or product matches over generic titles.
- If multiple links look useful, start with the closest match rather than loading everything.
- Ignore links with vague titles unless the task is also vague.
Fetch rules
- Resolve relative
hrefvalues against the current page URL. - Treat the linked file as the source of truth for browser-specific workflow.
- If the file cannot be fetched, note that and fall back to standard browsing.
Browser MCP guidance
- Use Browser MCP or the active browser tool to inspect
<head>content first. - Only switch into the linked skill workflow after confirming the title is relevant.
- Keep the fallback simple: no relevant link means no skill handoff.
Output expectation
After discovering a relevant web-skill, continue the task using that downloaded SKILL.md as the active operating guide.