second-brain: Skill for Claude Code

Jina Capture

Turn any URL into a stamped markdown file in wiki/raw/ using the Jina Reader service (https://r.jina.ai/<url>), which returns a clean markdown version of the page.

When to use

• User provides a URL and wants it captured into the second brain.
• User says "save this", "capture this", "add to raw", "read this later", etc., alongside a link.
• Bulk import of a list of URLs.

How it works

https://r.jina.ai/<full-url> returns the target page rendered as markdown. No API key is required for basic use. Just fetch it and save the body.

Steps

1. Confirm target folder. Default is wiki/raw/ in the current project (the second brain repo). If the user is not in that repo, ask where to save.

2. Fetch the markdown. For each URL:

   • Construct https://r.jina.ai/<url> (do not URL-encode the target — Jina expects the raw URL appended).
   • Use the WebFetch tool to retrieve it. Ask for the full markdown body.
   • If WebFetch summarizes instead of returning raw content, fall back to:

     curl -sL "https://r.jina.ai/<url>" -o /tmp/jina.md
     

     and read /tmp/jina.md.

3. Extract a title and slug. Jina's output usually starts with Title: ... and URL Source: ... lines followed by Markdown Content:. Parse the title from there; slugify it to max 50 chars, lowercase, hyphen-separated.

4. Write the file. Save to wiki/raw/<YYYY-MM-DD>-<slug>.md with frontmatter prepended:

   ---
   id: <YYYY-MM-DD>-<slug>
   date: <YYYY-MM-DD>
   source: <original url>
   tags: []
   status: draft
   ---
   

   Followed by the markdown body Jina returned. Keep the body intact — do NOT summarize or edit it. The point of raw/ is fidelity.

5. Report back. Print the new file path (as a computer:// link) and a one-line summary of what the article is about. Do not run /ingest automatically — the user decides when to distill.

Rules

• Never overwrite an existing file in wiki/raw/. If the slug collides, append -2, -3, etc.
• Never modify or delete other files in wiki/raw/.
• If Jina returns an error page or empty body, tell the user and do not create the file.
• For multiple URLs, process them in sequence and print one summary line per saved file at the end.