yichen-web-research

Route multi-stage internet research through search, archive, bookmark, and ASR child skills.

2.0k|274|Updated Feb 11, 2026
One-click install
npx skills add https://github.com/mcncarl/yichen-skills --skill yichen-web-research
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: yichen-web-research
Source: https://github.com/mcncarl/yichen-skills/tree/main/yichen-web-research
Command: npx skills add https://github.com/mcncarl/yichen-skills --skill yichen-web-research

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Deep internet research often spans searching, verifying original sources, archiving links, and transcribing media, but mixing these stages in one step leads to unauthorized downloads, unverified claims, and lost evidence trails. This Skill acts as a single safety-first router that plans the research, then delegates each stage to the correct child Skill with explicit authorization gates.

Core Features & Use Cases

  • Multi-Stage Routing: Directs requests to yichen-unified-search, yichen-content-archive, yichen-bookmarks-export, or yichen-asr based on task stage, never auto-archiving after a search.
  • Horizontal-Vertical Research Protocol: Builds a dated research brief, generates bounded longitudinal and cross-sectional query workstreams via an offline planner, and assembles a claim-source evidence ledger with structural stop gates before any report is written.
  • Backend Health Checks: Ships a read-only doctor script that verifies optional backends (AnySearch, Firecrawl, Zhihu CLI, Grok CLI, OpenCLI, ASR) without probing networks or reading credentials.
  • Use Case: Ask for a deep analysis of a company's development history plus its current competitive landscape; the Skill plans dated workstreams, searches per region and language, verifies claims against opened original sources, and only produces a report when all evidence gates pass.

Quick Start

Ask the assistant to run a horizontal-and-vertical deep research analysis on a company or industry with a defined scope, and let it plan and route the work through this Skill.

Frequently Asked Questions about yichen-web-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a deep research analysis with history and current landscape comparison?

Invoke the router with a research goal; it builds a dated brief, runs the offline plan_hengzong_research.py planner to create bounded longitudinal and cross-sectional workstreams, then searches, verifies original sources, and assembles evidence before writing any report.

When should I use the research router instead of a single search skill?

Use the router only when a request spans two or more stages, such as search plus archiving or transcription, or requires evidence-grounded longitudinal and cross-sectional synthesis. Single-stage searches, known-link reading, bookmark exports, or media transcription go directly to the corresponding child skill.

Does the research workflow automatically archive or download found content?

No. Search results remain candidates and never trigger automatic archiving or downloads. Archiving through the content-archive child skill requires an explicit user request with a bounded scope in the current turn.

What backends does the web research family support?

Optional backends include AnySearch for public search, Firecrawl for bounded Map and Scrape, a Zhihu CLI adapter for public search and hot lists, Grok CLI for native X search, OpenCLI adapters, and Step or Doubao ASR. Missing backends reduce coverage but never relax authorization rules.

Why does the research report stop with a blocking status?

Blocking occurs when structural gates fail, such as incomplete scope coverage, missing per-workstream verified claims, empty timelines or cross-sectional matrices, unresolved contradictions, or evidence links lacking locator, event date, or scope. Gaps need at least two distinct structured search attempts before disclosure.

Can the skill access private bookmarks or logged-in platform data?

Private bookmarks, feeds, and account data require explicit current-turn authorization for the specific platform and scope, and that authorization never transfers to media downloads. Bounded read-only Xiaohongshu and Douyin searches may reuse an existing Chrome session under fixed rate limits.