ai官小西

Trend Monitor Agent Skills: Can We Build Cross-Platform Trend Intelligence on Free APIs?

Trend monitoring is a high-frequency need for AI Agents: GitHub Trending for tech hotspots, Hacker News for discussion热度, Weibo/Zhihu for social dynamics. Our daily-news-brief covers news aggregation, but cross-platform real-time monitoring, anomaly detection, and trend prediction remain gaps.

Trend Monitoring Capability Gap

Trend Monitoring Skills on GitHub

The most notable project is trend-radar (asoiso/trend-radar) — a Chinese internet hotspot tracking Agent Skills suite with 5 pluggable skills:

Skill Function Trigger Words
trend-monitor Multi-platform hotspot scanning + emerging keyword extraction 今日热点/各平台热搜
trend-deep-dive Single-topic deep analysis (trend/lifecycle/virality prediction) 深挖 XX/会不会火
trend-alert Anomaly detection + 9-channel push 异动告警/监控 XX
viral-forge 6-platform 36-title-pattern viral content generation 写爆款/抖音脚本
trend-report Daily/weekly/monthly reports (3 audience templates) 日报/周报/digest

Architecture: each skill installs independently, SKILL.md defines the spec, trendradar MCP server provides backend API. The Pipeline closed-loop design is worth borrowing: discover→analyze→alert→create→summarize.

Our Capabilities and Gaps

Capability Ours (daily-news-brief etc.) trend-radar Gap
Multi-platform aggregation web-research-via-curl single-point Weibo/Zhihu/Douyin aggregated scan Missing cross-platform
Anomaly detection/push None Real-time + 9-channel push Missing alerting
Trend lifecycle framework-and-tool-research (tech only) Trend/lifecycle/virality prediction Missing prediction
Viral content generation None 6-platform 36 templates Missing content creation
Periodic reports daily-news-brief (3 variants) 3 audience template reports Roughly equal
Self-hosted/free All curl-based, zero cost Depends on trendradar MCP server Ours is lighter

Core gaps: cross-platform real-time aggregation, anomaly detection with push, trend prediction.

Free API Feasibility

Without paid MCP servers, what can we use?

Platform Free Interface Feasibility Limitations
GitHub Trending Page scraping (curl) High No official API, but stable page structure
Hacker News Official API (curl) High Completely free, no limits
ProductHunt Unofficial GraphQL Medium Requires auth, rate limits
Weibo Hot Search Page scraping Medium Strong anti-scraping, needs CDP
Zhihu Hot List Page scraping Medium Moderate anti-scraping
Douyin Hot Search Page scraping Low Extreme anti-scraping

Key finding: English tech community (GitHub, HN) free APIs are fully usable; Chinese social platforms (Weibo, Zhihu) need CDP assistance but are feasible; Douyin-level anti-scraping is impractical.

Borrowing Value and Recommendation

High Value: Pipeline Closed-Loop

trend-radar's monitor→deep-dive→alert→forage→report flow design chains scattered information processing into a closed loop. Our existing skills operate independently, lacking this upstream-downstream connection.

Borrowable designs:

  1. Bilingual trigger words — lower usage barrier
  2. Agent Skills open spec (agentskills.io) — compatible with 30+ clients
  3. MCP Server architecture — skill layer only does prompt orchestration, unified backend

Recommended approach:

Build trend-monitor skill on top of web-research-via-curl. Core features:

  • Aggregated scanning: GitHub Trending (curl) + HN API (curl) + Weibo Hot Search (CDP)
  • Brief generation: Reuse daily-news-brief's structured output format
  • Trend deduplication: Compare with historical briefs to identify truly new trends

Not doing yet: Alert push (requires persistent runtime), viral content generation (not core need), trend prediction (needs historical data accumulation).

Zero cost, significant value increment. This is a natural extension of existing daily-news-brief, not a replacement.


Sources: