last30days-skill-cn
last30days-skill-cn is a Chinese-language Agent Skill for researching recent public content across major Chinese internet platforms. It combines APIs, browser automation, public interfaces, and search fallbacks, then can emit compact, JSON, Markdown, context, path, or HTML output. It is a general Agent Skill rather than a DSH-only package.
Source repository: Jesseovo/last30days-skill-cn
Discover research workflow assets Explore EvoMap connectors
A skill focused on activity from the last 30 days
The default window is recent activity, but the CLI also supports a custom day count and an --as-of end date. Coverage can include Weibo, Xiaohongshu, Bilibili, Zhihu, Douyin, WeChat, Baidu, and Toutiao, subject to configuration and platform availability.
Example inputs:
Research how Chinese developer communities discussed local AI agents during the last 30 days.
Compare recurring concerns about AI browser tools across Bilibili, Zhihu, and Xiaohongshu.
How the 30-Day Window Gets Summarized
The tool can prefer configured APIs, fall back to Playwright-based browser collection, use public interfaces, and finally use public search results for some sources. That hierarchy improves resilience, but it does not guarantee complete coverage or precise engagement data.
Platform interfaces and access rules change. Use reasonable request frequency, respect terms and robots rules, and do not treat an empty result as proof that no discussion exists.
For browser-based collection infrastructure, review BrowserSkill for visible collection. For persistent research memory, see mem9.
Chinese-Language Output for Chinese-Speaking Teams
Use a source-first workflow:
- Define the topic, window, and target platforms.
- Diagnose which sources are currently available.
- Run the search with a narrow source set first.
- Review source links, dates, and fallback notes.
- Summarize only claims supported by the collected evidence.
Example with an output shape:
Return a Chinese report with: topic clusters, dated source links, disagreements, missing platforms, and three follow-up questions.
The project can generate an offline HTML report. That format improves review and archiving, but it does not validate the underlying claims by itself.
Installing last30days-skill-cn and Running Your First Summary
Install the Agent Skill with:
npx skills add Jesseovo/last30days-skill-cn -g
Optional browser support uses Playwright:
python -m pip install playwright
python -m playwright install chromium
Follow the last30days-skill-cn repository guide, then enter the installed Skill root before running its relative scripts/ paths:
cd <skill-root>
python scripts/last30days.py --diagnose
python scripts/last30days.py "AI编程助手" --emit compact
Use the host's documented Skills path when integrating with DeepSeek Harness. Agents with Bash, Read, and Write tools can use the Skill; DeepSeek Harness integration relies on its Skills path rather than a dedicated DSH plugin command.
Integration status on 2026-09-28: the official DSH project is still under developer preview and warns of compatibility-breaking changes. Confirm the Skills directory after a host upgrade.
Preserve a Verified Research Pattern With EvoMap
This Skill performs research; EvoMap is separate infrastructure for organizing verified agent assets. When the sources, window, fallbacks, and gaps have been validated, a team can explicitly package the method for reuse. Start from the EvoMap Wiki and use Vibe-Skills for multi-Skill routing.
FAQ
Does every platform work without configuration?
No. Availability differs by source, browser setup, credentials, login state, and platform behavior. BrowserSkill can support visible browser collection when that route is appropriate.
Is the window fixed at exactly 30 days?
No. The CLI supports --days and --as-of options for a different window. Record the chosen period in the output so readers can interpret the results.
Can collected content be used without checking platform rules?
No. Users must respect applicable law, platform terms, and robots rules. Keep source links and collection notes with the final summary.
Reuse a Traceable Research Method
Package a traceable research method only after its sources, date window, fallbacks, and missing coverage have been recorded.