Custom Keyword Public Sentiment Clean

Purpose

This skill collects and summarizes public content from four platforms for a user-defined keyword:

• Xiaohongshu
• Douyin
• Weibo
• Bilibili

It is intended for quick cross-platform monitoring of games, products, campaigns, brands, or public topics within a recent 24-hour window.

Inputs

• keyword: primary keyword, required
• aliases: optional aliases, comma separated
• strong_keywords: optional strong-related terms, comma separated
• workdir: full Windows output directory, required

Standard workflow

1. Run a headless full scan first:

python scripts/cloak_scrape.py <workdir> --platform all --headless --keyword "<keyword>" --aliases "alias1,alias2" --strong-keywords "term1,term2"

2. Read only === SCRAPE SUMMARY === and NEED_LOGIN.
3. Rerun only blocked platforms in headful mode if needed.
4. Aggregate the report:

python scripts/aggregate.py <workdir>

Outputs

• query_config.json
• xhs_raw.txt
• douyin_raw.txt
• weibo_raw.txt
• bilibili_raw.txt
• <keyword_slug>_24h_report.md
• evidence screenshots for sorting, captcha, and page-state verification

Design principles

• Headless full scan first, then targeted headful reruns only when needed
• Use SCRAPE SUMMARY as the main control signal
• Use query_config.json for dynamic filtering and report naming
• Reuse verified sorting and extraction logic per platform where available

Beta note

• This is a Beta release.
• The Bilibili path has passed smoke testing.
• Xiaohongshu, Weibo, and Douyin may require login or may trigger verification challenges.
• Recommended usage: run headless first, then rerun only blocked platforms in headful mode.
• See BETA_DELIVERY.md for current delivery scope and limitations.