User requests to fetch content from a URL.
Agent invokes the `agent-fetch` skill with the provided URL.
`agent-fetch` fetches the URL using browser impersonation and multiple extraction strategies.
The skill selects the best extraction result based on completeness.
Metadata (title, author, date) is composed from the best source.
The extracted content is formatted as markdown.
The agent presents the formatted content to the user.
If the fetch fails, the agent suggests a retry or informs the user.