Domestic Self-Media Hotspot Extractor

Problem

Domestic social media hot lists are fragmented across platforms and often unstable: some endpoints require login, some return 0 items, and some lack cover images. Using raw trending lists directly makes it hard to remove duplicates, verify sources, merge cross-platform signals, and separate weak topics from actionable ones. This skill treats hotspot extraction as an auditable collection pipeline rather than one-off scraping.

How It Works

• Multi-source collection: Covers weibo, bilibili, baidu, douyin, toutiao, xiaohongshu, and more, with explicit primary and fallback paths plus image-field status.
• Failure handling: When a source returns login_required, a CAPTCHA, or zero rows, it does not silently skip the source; it tries backup APIs, Chrome bridge collection, and limited candidate sources while preserving failure evidence in the report.
• Normalization and filtering: Uses normalizer, deduplicator, and keyword_filter to unify fields, merge similar topics, and remove sensitive or promotional terms.
• Scoring and quality gate: Scores topics by heat growth, cross-platform resonance, confidence, content fit, and timing, then checks trusted sources, timeliness, and degraded-source ratio before delivery.
• Image extraction: Can extract page images for YouTube, TikTok, Xiaohongshu, and others via CDP or AppleScript, but depends on local Chrome debugging ports and user permissions.

Caveats

Best for turning hot topics into a content queue, editorial calendar, or monitoring report. Not ideal for sub-second real-time updates or links that require guaranteed login-free collection from every platform. The output is a status-marked JSON file; login-required sources should be resolved by a human before downstream use.