What problem does it solve?
Searching the public web and multiple social platforms usually means juggling many CLIs, APIs, and login states with inconsistent output formats. This Skill plans one bounded, read-only route per query and normalizes results into a single reviewable candidate schema.
Core Features & Use Cases
- Offline route planning:
route_search.py selects one bounded backend (AnySearch, AI HOT, Firecrawl, Zhihu, Weibo, YouTube, X/Grok, and more) and emits explicit steps, authorization state, and limitations without calling any service.
- Unified candidate envelopes: Bundled adapters output schema 1.0 with candidates, coverage, provenance, and sanitized errors, plus short-lived HMAC receipts for verifying current-search candidates.
- Platform coverage: Discovers public content on GitHub, Zhihu, WeChat Official Accounts, Weibo, Xiaohongshu, Douyin, Toutiao, X/Twitter, Bilibili, YouTube, and Xiaoyuzhou, with strict read-only and login-state boundaries.
- Use Case: Ask for recent public discussion of a product across Weibo and X; the planner emits bounded steps, adapters return normalized candidates, and you verify only the items you intend to cite.
Quick Start
Ask the assistant to search a topic on a specific platform with a time range and result limit, for example: search Weibo for public posts about iPhone 18 from the last 30 days and return up to 20 candidates.
