Research What the Chinese Internet Said in the Last 30 Days with last30days-cn

last30days-cn is Jesse's 2k-star MIT Chinese localization of Matt Van Horn's last30days skill: eight Chinese platforms by default, engagement-weighted ranking, and every claim tied back to its source.

Why This Skill Matters

Per the README, last30days-cn searches recent discussions, ranks them by relevance, recency, and engagement, and groups the same event across platforms while keeping the original sources. The default eight are Weibo, Xiaohongshu (RED), Bilibili, Zhihu, Douyin, WeChat public accounts, Baidu, and Toutiao; you can add Hacker News, GitHub, and Reddit, or connect an installed upstream skill for more overseas sources. Output comes as compact agent evidence, Markdown, JSON, and offline HTML, and reports state which sources returned results, how they were retrieved, and where coverage is missing.

It also ships a second mode: a daily cross-platform hot board, collected automatically at 08:00, 12:00, and 20:00 Beijing time, that brings together Chinese hot lists, tech news, and Hacker News. SkillMap covered the upstream skill in an earlier tutorial.

Installation

The README's install path is a sentence for your agent. It works with agents that can run shell commands, including Claude Code, Codex, Cursor, and Gemini CLI; the agent handles installation, environment checks, and the first run, and asks you to complete any required platform login.

Install and configure https://github.com/Jesseovo/last30days-skill-cn, check which sources are available, and use it to show me today's cross-platform hot board.

Real Workflow: Topic Research and the Hot Board

The README's example requests cover both modes — a hot-board snapshot of today, a seven-day topic study, and a widening to overseas sources:

What is trending in AI today? Pick three stories worth a closer look.

Research the last seven days of discussion about AI coding assistants in China. Compare views across platforms and cite the sources.

Add voices from Hacker News, GitHub and Reddit, and generate an offline report.

Under the hood, the skill's SKILL.md documents a single script with flags. Topic research runs all eight Chinese platforms unless you narrow it:

python {{SKILL_DIR}}/scripts/last30days.py "AI 编程助手" --emit compact
python {{SKILL_DIR}}/scripts/last30days.py "AI 编程助手" --deep --emit compact        # more pages per platform
python {{SKILL_DIR}}/scripts/last30days.py "AI 编程助手" --search bilibili,zhihu,xhs --emit compact
python {{SKILL_DIR}}/scripts/last30days.py "AI 编程助手" --days 7 --emit compact
python {{SKILL_DIR}}/scripts/last30days.py "AI 编程助手" --emit html-path            # offline HTML report path

The hot board needs no topic and fetches the Weibo, Baidu, Douyin, Toutiao, Bilibili, and Zhihu boards in parallel; --hot-sources adds news (IT Home, Huxiu, SSPAI, iFanr, Solidot) and global (the Hacker News front page):

python {{SKILL_DIR}}/scripts/last30days.py --hot --emit compact
python {{SKILL_DIR}}/scripts/last30days.py --hot "AI" --emit compact
python {{SKILL_DIR}}/scripts/last30days.py --hot --hot-sources boards,news,global --emit compact

The skill instructs the agent to lead with cross-platform clusters (跨平台热点) — the same event on several boards at once is the strongest signal — and to say that hot lists are real-time rankings, not 30-day totals.

Tips

• The query type (breaking news, how-to, product, opinion, comparison, prediction, concept) is detected automatically; it changes ranking tie-breaks and, in --quick mode, which platforms run.
• Overseas engines do not match Chinese text: for a Chinese topic, the SKILL.md tells the agent to translate it into concise English and pass it as --global-query. X, YouTube, TikTok, and Instagram go through a bridge to an installed mvanhorn/last30days skill, which needs its own keys and Python 3.12 or newer.
• Since 2025, per the SKILL.md, Weibo, Xiaohongshu, Zhihu, and Douyin search mostly require a logged-in session; without one the engine falls back to hot lists and public web search. python {{SKILL_DIR}}/scripts/last30days.py login xiaohongshu opens a visible browser for you to scan a QR code or sign in, --cookie accepts a cookie from your own browser on machines without a display, and --diagnose prints which platforms work.
• Results are cached for 24 hours; --refresh ignores the cache and --no-cache skips reading and writing it. Settings live in ~/.config/last30days-cn/.env or a project's .claude/last30days-cn.env, and everything is optional.

When Not to Use This

Login requirements and upstream availability affect coverage: reports distinguish logged-in results, hot-list matches, and public-search fallbacks, and public-search results have no reliable engagement counts or exact dates — the README says they do not represent a complete platform sample. The SKILL.md scopes the skill to learning, research, and personal knowledge work, asks you to keep request frequency low and respect platform terms and robots.txt, and rules out large-scale scraping, personal-data collection, and commercial data services.

See the leaderboard for more skills.