Agent Skills
› MarsZ42/OrbitOS
› ai-products
ai-products
GitHub从多源抓取、去重并排名AI新产品发布，生成包含内容角度的摘要报告。支持手动调用及每日启动场景，具备缓存和容错机制。
触发场景
用户输入 /ai-products
触发 /start-my-day 流程
安装
npx skills add MarsZ42/OrbitOS --skill ai-products -g -y
SKILL.md
Frontmatter
{
"name": "ai-products",
"description": "Curate AI product launches from Product Hunt, Hacker News, GitHub, and Techmeme. Use when user invokes \/ai-products or when \/start-my-day needs product launches."
}
AI Product Discovery
Fetch, deduplicate, and rank AI product launches from multiple sources.
Sources
| Source | URL | Notes |
|---|---|---|
| Product Hunt | https://www.producthunt.com/feed |
Filter for AI-related |
| Hacker News | https://hn.algolia.com/api/v1/search?tags=show_hn&numericFilters=created_at_i>TIMESTAMP |
Show HN posts, 24h window |
| GitHub Trending | https://mshibanami.github.io/GitHubTrendingRSS/daily/python.xml |
Python repos |
| Techmeme | https://techmeme.com/river |
Product announcements |
Workflow
-
Check cache: Look for
50_Resources/ProductLaunches/YYYY-MM/YYYY-MM-DD-Digest.md. If exists with today's date, return cached. -
Fetch sources: Use WebFetch on each. Extract product name, URL, description, and engagement metrics (votes/points/stars).
-
Filter: Keep only AI-related products (keywords: AI, ML, LLM, GPT, Claude, automation, agent, model).
-
Deduplicate: Same product across sources = merge. Keep best description, combine metrics, track all sources.
-
Rank by:
- AI relevance
- Engagement (normalize: PH votes/500, HN points/100, GH stars/1000)
- Content potential (tutorial-friendly, review-worthy, open source bonus)
- Recency and novelty
-
Generate digest: See TEMPLATE.md. Sections:
- Top Picks (3-5) with content angles
- LLM & AI Models
- Developer Tools
- Productivity & Automation
- Open Source Highlights
-
Save files:
50_Resources/ProductLaunches/YYYY-MM/YYYY-MM-DD-Digest.md50_Resources/ProductLaunches/YYYY-MM/Raw/YYYY-MM-DD_ProductHunt-Raw.md50_Resources/ProductLaunches/YYYY-MM/Raw/YYYY-MM-DD_HackerNews-Raw.md50_Resources/ProductLaunches/YYYY-MM/Raw/YYYY-MM-DD_GitHub-Raw.md
Output Format
Manual invocation: Full digest with all sections.
From /start-my-day: Condensed list:
**Product Launch Opportunities (5):**
- [Product] - [Angle] - [Top metric]
...
Full digest: [[YYYY-MM-DD-Digest]]
Error Handling
- Source down: Continue with others, note in digest
- <2 sources available: Fall back to yesterday's archive
- Empty results: Create minimal digest noting "No new AI products"
Content Angle Logic
- High engagement + tutorial-friendly: "Tutorial opportunity"
- Novel + early stage: "First-mover advantage"
- Open source + complex: "Deep dive analysis"
- SaaS + practical: "Tool review"
- Similar to existing: "Comparison vs [competitor]"
版本历史
- 921bd14 当前 2026-07-25 10:48
