README 宣告 MIT 授權的 Agent 技能，對 8 大中文平台執行 AI 網路搜尋並產出近 30 天有據可查的研究報告；可作為通用技能接入 DeepSeek Harness（代理需具備 Bash / Read / Write 工具）。
DSH 整合
相容
作者聲明
安全稽核
未稽核
最後核實
2026-08-21
授權條款
NOASSERTION
01它能幫你完成什麼？
Research what happened on China's 8 major internet platforms in the last 30 days for any topic
A cited research report built from deduplicated, weighted-scored (45% relevance / 25% freshness / 30% engagement) results across Weibo, Xiaohongshu, Bilibili, Zhihu, Douyin, WeChat, Baidu and Toutiao — emitted as compact / JSON / Markdown / context / offline HTML
Researchers, analysts and content teams who need current, well-sourced insight into the Chinese internet straight from their AI agent
Turn multi-platform Chinese web search results into a browsable, archivable offline HTML research report
An offline-openable `report.html` (Swiss/IKB visual language from guizang-ppt-skill) generated via `--emit html` / `--emit html-path`
Users who want a printable, shareable research deliverable instead of terminal output
02如何將外掛接入 DeepSeek Harness？
先決條件
- A Python runtime — the skill runs as a Python CLI (`scripts/last30days.py` / `skills/last30days/scripts/last30days.py`)
- An AI agent exposing Bash / Read / Write tools — the README states any such agent can use this skill
- Optional: Playwright (`pip install playwright` + `playwright install chromium`) to unlock key-free crawler mode on 7 of 8 platforms
- Optional: `jieba` for enhanced Chinese tokenization (CJK bigram fallback runs without it)
安裝步驟
- 01
Install via Agent Skills (recommended): `npx skills add Jesseovo/last30days-skill-cn -g`
$ npx skills add Jesseovo/last30days-skill-cn -g
- 02
Or clone the repo into your agent's skills directory, e.g. `git clone https://github.com/Jesseovo/last30days-skill-cn.git ~/.claude/skills/last30days-cn`
$ git clone https://github.com/Jesseovo/last30days-skill-cn.git ~/.claude/skills/last30days-cn
- 03
Optional (recommended): `python -m pip install playwright` then `python -m playwright install chromium` for crawler mode
- 04
Optional: `python -m pip install jieba` for better Chinese tokenization
- 05
Optional: create `~/.config/last30days-cn/.env` (chmod 600) and fill in optional API keys — only WeChat search requires `WECHAT_API_KEY`
驗證整合成功
- Run `python scripts/last30days.py --diagnose` — it prints each platform's availability status and the crawler engine status as JSON
03DSH 整合程度與能力邊界
Generic Agent Skill (SKILL.md + Python CLI) installable into any AI agent that exposes Bash / Read / Write tools — DeepSeek Harness included. The README does not document a dsh-specific integration; it documents Agent Skills / Claude Code / Cursor / OpenClaw installs plus a generic-agent mode, and the repo is tagged `dsh-plugin`.
30-day cross-platform research on 8 Chinese platforms
a natural-language topic plus optional flags (`--quick` / `--deep` / `--days N` / `--as-of` / `--search <platforms>`)→deduplicated, weighted-scored results from Weibo, Xiaohongshu, Bilibili, Zhihu, Douyin, WeChat, Baidu and Toutiao rendered as compact / JSON / Markdown / context report
performs live HTTP requests to the target platforms (and Bing site: search fallback)`--save-dir` writes raw outputs to disk; `--emit html` / `html-path` writes a `report.html` fileThree-tier data-acquisition fallback with optional Playwright crawler
per-platform search requests→API mode → Playwright crawler mode → public endpoints → Bing site: search fallback, so 7 of 8 platforms work with zero API keys once Playwright is installed
crawler mode drives a real Chromium browser via Playwright and simulates user browsingcrawler mode persists login cookies (login-state cache) between runsWeighted result scoring
raw multi-platform search results→a 0–100 composite score per item: 45% relevance, 25% freshness, 30% platform engagement (likes / comments / shares / views etc.)
04適合誰？何時不該用？
適合
- Researchers, analysts and content teams who need current, well-sourced insight into the Chinese internet straight from their AI agent
- Users who want a printable, shareable research deliverable instead of terminal output
不適合
- The project is declared for learning and research purposes only and explicitly forbids commercial use; users must comply with Chinese laws (Cybersecurity / Data Security / Personal Information Protection), platform ToS and robots.txt, and bear all legal risk of crawling.
- Platform interfaces can change at any time and availability is not guaranteed; Douyin / Toutiao native web APIs require signed parameters (`a_bogus` / `_signature`) and are often risk-controlled, so their fallback only yields public links without real engagement data or exact dates.
- Aggressive crawling risks platform bans; the README advises keeping request frequency reasonable (e.g. ≥ 5 seconds between searches).
05相容性、維護與安全提醒
- The project is declared for learning and research purposes only and explicitly forbids commercial use; users must comply with Chinese laws (Cybersecurity / Data Security / Personal Information Protection), platform ToS and robots.txt, and bear all legal risk of crawling.
- Platform interfaces can change at any time and availability is not guaranteed; Douyin / Toutiao native web APIs require signed parameters (`a_bogus` / `_signature`) and are often risk-controlled, so their fallback only yields public links without real engagement data or exact dates.
- Aggressive crawling risks platform bans; the README advises keeping request frequency reasonable (e.g. ≥ 5 seconds between searches).
MIT per README (GitHub reports NOASSERTION) · actively maintained (latest release v3.2.0, 2026-07-20)
06常見問題
如何在 DeepSeek Harness 中使用 last30days-cn？
它是通用的 Agent 技能：README 寫明任何支援 Bash / Read / Write 工具的 AI Agent 都可使用。用 `npx skills add Jesseovo/last30days-skill-cn -g` 安裝，或把儲存庫複製到代理的技能目錄，接著讓代理執行 `python scripts/last30days.py "<主題>"` 即可。README 未提供針對 dsh 的專屬整合說明。
是原生整合還是透過 MCP？
兩者皆非——它是由 SKILL.md 驅動的 Python CLI 技能，不是 MCP 伺服器，也不是 dsh 原生執行環境。代理讀取 SKILL.md 後透過 Bash 工具呼叫命令列入口。
需要設定 API Key 嗎？
所有 API Key 都是選配。安裝 Playwright 後，7/8 的平台可透過爬蟲模式免 Key 使用；只有微信公眾號搜尋需要 `WECHAT_API_KEY`。`jieba` 也是選配，未安裝時會自動退回 CJK bigram 分詞。
會有哪些網路請求與副作用？
它會對 8 大中文平台發出即時 HTTP 請求（可選 Playwright 瀏覽器自動化，並持久化登入 Cookie），最後一級備援是 Bing site: 站內搜尋。README 建議每次搜尋間隔至少 5 秒以避免被平台封鎖，且本專案僅限非商業的學習研究用途。
如何驗證安裝成功？
執行 `python scripts/last30days.py --diagnose`，它會輸出各平台的可用狀態與爬蟲引擎狀態的 JSON 診斷結果。
07相關的 DSH 工作流程
browser4
作者 platonai
面向自主智慧體與大規模網頁自動化的 AI 原生瀏覽器引擎，支援智慧內容抽取。
dsh-browser
作者 lum1104
Chrome sidebar extension that lets DeepSeek Harness operate your browser directly, no vision capabilities required. 一款 Chrome 側邊欄擴充套件程式，可讓 DeepSeek Harness 直接操控您的瀏覽器，無需視覺能力。
modsearch
作者 liustack
🥇 The strongest free web search plugin for DeepSeek Harness, and the search bridge for every model without native web access. Free, no signup, no API key. Ask the web or X, get structured JSON evidence. | 🥇 全網最強的 DeepSeek Harness 免費聯網搜尋外掛，免費免註冊免 API key。為不能聯網的模型補上搜索，問網頁或 X，拿回結構化 JSON 證據（搜尋、抓取、引用）。
anysearch-dsh
作者 anysearch-team
為 DeepSeek Harness 提供的 AnySearch 網頁搜尋提供方與高階搜尋工具。
08資料與來源
### 通用 Agent 任何支持 **Bash / Read / Write** 工具的 AI Agent 都可以使用本技能。
**last30days-cn** 是一个 AI Agent 技能（Skill），能够自动搜索中国互联网 8 大主流平台最近 30 天的内容，综合分析后生成有据可查的研究报告。
此頁面根據專案公開文件、儲存庫中繼資料與 DSH Plugins 的結構化解析所產生；最後核實於 2026-08-21。發現錯誤？提交更正。
最佳 DeepSeek Harness 外掛
從全目錄挑出的 12 個值得優先安裝的外掛，涵蓋各個分類。