news-briefing - SKILL.md Agent Skill

name: news-briefing description: "Synthesize a daily news briefing JSON document from a pre-fetched raw briefing file."

Daily News Briefing

Overview

The deterministic crawl/extract pipeline now lives in the briefings repo.

Use this skill only after a raw file already exists at:

~/Code/HerbCaudill/briefings/public/briefings/raw/YYYY-MM-DD.json

Do not crawl source sites, fetch headlines, or extract article bodies here. The raw file already contains the candidate articles, source metadata, first-seen positions, sightings, and extracted body text.

Input

Read the target raw file and synthesize a final briefing JSON document for the app.

Output shape

Return a single JSON object:

{
  "sections": [
    {
      "title": "World",
      "stories": [
        {
          "headline": "Story headline",
          "body": "Details and context based on article content, one paragraph, no markdown.",
          "sources": [
            { "name": "BBC News", "url": "https://www.bbc.com/news/articles/abc123" },
            { "name": "Al Jazeera", "url": "https://www.aljazeera.com/news/2026/..." }
          ]
        }
      ]
    }
  ]
}

Section titles in order: World, US, Spain, Barcelona & Catalunya.

Rules

Use the raw article bodies as the factual basis for summaries
3-5 stories per section when enough source material exists
Prioritize stories covered by 2+ sources when possible
Exclude sports entirely
Each story body must be one paragraph of plain text with no markdown or links
sources[] must point to the actual article URLs from the raw file
Within each section, order by strength of coverage and significance
No emojis anywhere in the output
Output only the JSON object — no surrounding text or code fences