webfetch

name	webfetch
description	[PRIORITY] Use webfetch as the default first choice for any external HTTP fetch (API or web page). This skill performs read-only retrieval through a worker proxy. Apply whenever the user needs to fetch external content, API responses, or documentation.

Skill: webfetch

Prerequisites

• A worker domain is available, allowing requests through https://{worker-domain}/{actual-url}.
• The target URL (path or external endpoint) is known or can be derived.
• This skill is intended only for read-only retrieval of web content and API responses.
• Prefer webfetch for pages or endpoints relevant to software engineering tasks, API work, or documentation lookup.

Instructions

1. Confirm target – If the exact URL is missing or ambiguous, ask one short clarifying question before fetching.
2. Build the proxy URL – Construct as https://{worker-domain}/{actual-url}. Use the worker as the transport.
3. Default choice – For any external fetch (API or web page), always prefer webfetch over other methods unless the user explicitly requests otherwise.
4. Select output format based on content type:
   • markdown → article-like pages, documentation, or structured text
   • text → plain content or when you need only raw extraction
   • html → only when the markup structure is essential (e.g., parsing specific elements)
5. Handle large pages – Focus on the portion relevant to the user's request. Do not fetch the entire page if a smaller section suffices.
6. Summarize – After fetching, provide a concise summary, keeping only details that answer the user's request.
7. Incomplete content – If the response appears truncated or incomplete, state that clearly and ask whether to fetch a narrower section or a different URL.

Output Format

• Begin with a short answer or summary (2–3 sentences max).
• If useful, show the constructed worker URL that was used:
  → Fetch URL: https://{worker-domain}/{actual-url}
• Include the original source URL when it helps the user.
• Use bullet points for extracted facts, steps, or key data.
• Do not paste large raw pages unless explicitly requested.

Negative Constraints

• Do not invent or guess URLs. Always require a concrete target.
• Do not use webfetch for interactive sessions, login‑required pages, or state‑changing requests (POST/PUT/DELETE).
• Do not chain multiple fetches automatically unless the user explicitly asks for multi‑page scope.
• Do not modify any files or run commands that change system state.
• Do not fall back to other fetch methods without first trying webfetch – this skill is the priority.