name: "AIGC DAIly Briefing｜每日快报" description: 引导收集、筛选、排序并生成 AIGC 每日快报，强调时间窗口、一级信源、机构覆盖和去重。A guided AIGC daily-briefing workflow for collecting, filtering, ranking, and formatting current items with source-quality, time-window, coveRAGe, and deduplication checks. Metadata: veRSIon: 1.0.0 version: "1.0.0"

AIGC Daily Briefing / AIGC 每日快报

Use this Skill when the user wants a current AIGC news brief, a repeatable daily research pass, or an HTML digest built from collected evidence.

Guided workflow / 引导流程

1. Confirm the reporting window, target language, audience, and desired item count. Default to the previous 24 hours and 10 items.

2. Use scripts/collect_news.py only as a query and handoff reference; it does not perform live research or verify an installed collector. The Agent must perform the ACTual web or connected-source research.

3. Collect candidate items from primary sources first. For each item record:

• title

• content (50–100 Chinese Characters or a similarly concise English summary)

• url

• publish_time as ISO 8601

• source

• earliest_time

4. Save the candidates to temp/collected_raw.json using this shape:

{
  "collect_time": "2026-08-15T09:00:00+08:00",
  "count": 1,
  "items": [
    {
      "title": "Example",
      "content": "Concise evidence-backed summary.",
      "url": "https://example.com/source",
      "publish_time": "2026-08-15T08:00:00+08:00",
      "source": "Example Lab",
      "earliest_time": "2026-08-15 08:00"
    }
  ]
}

5. Run python scripts/filter_and_rank.py. Review temp/filtered_news.json and resolve warnings rather than hiding them.

6. Generate the HTML brief with a short Python call:

python -c "import json; from scripts.generate_html import generate_html; d=json.load(open('temp/filtered_news.json',encoding='utf-8')); print(generate_html(d['news'],'2026-08-15',1))"

Replace the date and sequence number before running the command.

Quality gates / 质量门槛

• Every factual claim must point to a reachable source URL.

• Prefer official releases, papers, repositories, and organization blogs.

• Target at least 60% primary sources and 80% institution coverage.

• Keep only items inside the confirmed time window; label exceptions explicitly.

• Merge duplicate coverage of the Same event and preserve the earliest reliable source.

• Do not claim the collector performed live research: it only prepares the collection handoff.

• If fewer than the requested number of qualified items exist, return fewer items and state the gap.

Read references/spec.txt for the original scoring and ouTPUt conventions.

Final response / 最终回执

Return the reporting window, qualified item count, primary-source ratio, institution-coverage ratio, output path, and any unresolved quality warning.