Skip to content

feat: add 13 Chinese Internet + AI Productivity Skills#335

Open
AaronAust1n wants to merge 1 commit into
OpenBMB:mainfrom
AaronAust1n:feat/china-productivity-skills
Open

feat: add 13 Chinese Internet + AI Productivity Skills#335
AaronAust1n wants to merge 1 commit into
OpenBMB:mainfrom
AaronAust1n:feat/china-productivity-skills

Conversation

@AaronAust1n

Copy link
Copy Markdown

PR: Add 13 Chinese Internet + AI Productivity Skills

Summary

This PR adds 13 ready-to-use Agent Skills that fill the gap in PilotDeck's skill ecosystem for Chinese-language and domestic-platform workflows.

Current state: PilotDeck ships with 26 skills (notion, obsidian, github, tmux, weather, etc.) — all targeting English-language / international platforms. There is zero coverage for China's major content platforms (WeChat, Zhihu, Bilibili, Douban, Juejin, Weibo) or Chinese-language AI research pipelines.

This PR adds:

  • 6 Chinese content retrieval skills (WeChat articles, Zhihu columns, Bilibili metadata, Douban movies/books, Juejin search, Weibo hot search)
  • 4 global AI/research monitoring skills (arXiv, Hacker News, GitHub trending, AI papers trending via OpenAlex)
  • 3 content production skills (HTML intelligence reports, podcast scriptwriting, VoxCPM TTS)

All skills use the same SKILL.md format as existing ones (weather, skill-creator, etc.) and work with zero additional installation (12/13 need only curl + python3; voxcpm-tts optionally needs pip install voxcpm).

Skills Added

Skill What it does Auth
wechat-mp-fetch Fetch WeChat Public Account article body None (MicroMessenger UA)
zhihu-answer-fetch Fetch Zhihu column articles (full) + answers (~2K) None
bilibili-video-info Video metadata + tags + danmaku XML None
douban-media-lookup Movie/book search + details (rating, cast, ISBN) API key (documented)
juejin-article-search Search + full-text extraction from SSR pages None
weibo-hot-search Real-time hot search (50 items + volumes + tags) None (Referer required)
github-trending Trending repos via Search API Optional token
arxiv-cn-daily Latest AI/ML papers by category None
hackernews-ai-trending HN AI stories via Algolia search None
ai-papers-trending Top-cited papers via OpenAlex + S2 fallback None
html-report-cn Self-contained Chinese HTML briefings N/A (prompt skill)
podcast-scriptwriter Two-host podcast script with tone annotations N/A (prompt skill)
voxcpm-tts Speech synthesis via VoxCPM (30 langs, cloning) N/A (open-source model)

Testing

All 13 skills were tested against live public APIs in 3 rounds, 65 test cases total:

  • Round 1: API connectivity verification (28 cases)
  • Round 2: Failure fix + boundary testing (31 cases, 27 edge cases)
  • Round 3: Independent retest of fixed paths + root-cause analysis (6 cases)

Final verdict: 8 GO / 5 GO_WITH_CAUTION / 0 NO_GO.

The 5 caution items each have limitations explicitly documented in their SKILL.md (truncation, weekend skipDays, leaked apikey risk, endpoint-specific quirks, GPU requirement).

Design Decisions

  1. Anthropic Skill format (SKILL.md + YAML frontmatter) — consistent with existing PilotDeck skills (weather, skill-creator, etc.)
  2. Zero external pip dependencies for 12/13 skills — only system curl + python3 standard library
  3. Every curl command is verified against live APIs — no hypothetical endpoints
  4. Each SKILL.md documents failure modes explicitly — what happens when things break, and how to recover
  5. Fallback chains where applicable — WeChat (MicroMessenger UA → WeWe-RSS), Weibo (hot_band → tophub), Douban (API → web scrape), GitHub (Search API → HTML), AI papers (OpenAlex → Semantic Scholar)
  6. Intentionally excluded: Xiaohongshu — aggressive anti-scraping + ¥4.9M court judgment (2025), no stable public path exists

Checklist

  • All SKILL.md files have valid YAML frontmatter (name + description)
  • All SKILL.md files are under 500 lines (max is voxcpm-tts at 213 lines)
  • Descriptions are third-person, include WHAT + WHEN trigger words
  • No external CSS/JS in html-report-cn template
  • Rate limits documented and honored in example code
  • No sensitive credentials (Douban apikey is public/leaked, documented as such)
  • Consistent terminology throughout
  • install.sh tested end-to-end

How to Test

cd skills/
# Test one skill (e.g., arxiv)
curl -s "https://export.arxiv.org/api/query?search_query=cat:cs.AI&sortBy=submittedDate&sortOrder=descending&max_results=3"
# Should return Atom XML with 3 paper entries

# Or run the full install
chmod +x install.sh && ./install.sh

Related

  • Closes: #XX (if there's a "Chinese skill support" issue)
  • Extends: existing skills/ directory
  • Compatible with: PilotDeck ≥ 0.1.0, any Anthropic-Skill-format runtime

Add 13 ready-to-use Agent Skills filling the gap in PilotDeck skill
ecosystem for Chinese-language and domestic-platform workflows:

Chinese content retrieval (6): wechat-mp-fetch, zhihu-answer-fetch,
bilibili-video-info, douban-media-lookup, juejin-article-search,
weibo-hot-search

AI/research monitoring (4): github-trending, arxiv-cn-daily,
hackernews-ai-trending, ai-papers-trending

Content production (3): html-report-cn, podcast-scriptwriter,
voxcpm-tts

All skills use Anthropic SKILL.md format. 12/13 need only curl +
python3 stdlib. 65 test cases across 3 rounds verified against
live APIs. Includes plugin.json manifest and README.

Tested: 8 GO / 5 GO_WITH_CAUTION / 0 NO_GO.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant