page-digest

Page Digest

Turn any URL into a concise, structured markdown digest.

When to use

• User shares a link and wants "the gist" or "summary"
• Need to extract action items, key facts, or entities from an article
• Quick briefing before a meeting based on a web page

Prerequisites

• web_fetch tool available in the agent runtime
• Target URL must be publicly accessible (no auth-gated pages)

Steps

1. Receive URL — Accept the target URL from the user or upstream process.
2. Fetch content — Call web_fetch with extractMode: "markdown" and maxChars: 20000.
3. Analyze — Read the fetched markdown and identify:
   • Title (first H1 or page title)
   • TL;DR (≤ 3 sentences)
   • Key points (3-7 bullet points)
   • Entities mentioned (people, orgs, products — max 5)
   • Action items (if any — explicit next steps or recommendations)
4. Format output — Produce a markdown digest with the sections above, in that order.
5. Return — Deliver the digest to the caller.

Output format

## {Title}

**TL;DR:** {3-sentence summary}

### Key Points
- {point 1}
- {point 2}
- ...

### Entities
- {entity} — {brief role/context}

### Action Items
- {action item 1}
- {action item 2}

Example

Input: https://example.com/article-about-ai-regulation

Output:

## AI Regulation Bill Passes Senate

**TL;DR:** The U.S. Senate passed a sweeping AI regulation bill Thursday. The legislation requires companies to audit high-risk AI systems before deployment. The bill now moves to the House for consideration.

### Key Points
- Bipartisan vote of 68-32
- Covers high-risk AI in healthcare, finance, and critical infrastructure
- Penalties for non-compliance up to $10M or 2x profits
- Enforcement begins 18 months after signing

### Entities
- U.S. Senate — Legislative body that voted
- Senate Majority Leader — Sponsored the bill
- FDA — Will enforce medical AI provisions

### Action Items
- Review compliance timeline for affected systems
- Schedule legal review of audit requirements

Error handling

• If URL is unreachable: report the error with the URL, suggest alternatives
• If content is too short (< 200 chars): note it and produce what's available
• If content is too long (> 20000 chars): truncate and warn that the digest may be incomplete