Agentic AI Atlas

II.

Page reference

page:docs-reference-repos-blessonism-openclaw-search-skills-research

Reading · 2 min

blessonism/openclaw-search-skills reference

OpenClaw-specific search skills providing multi-source parallel search, content extraction, and document parsing. Three composable skills: search-layer, content-extract, mineru-extract.

Pagewiki/docs/reference-repos/blessonism/openclaw-search-skills/research.mdOutgoing · 0Incoming · 1

blessonism/openclaw-search-skills

Metadata

**Stars:** 408
**License:** MIT
**Last pushed:** 2026-03-18
**Topics:** (none)
**Fork:** No

Archetype: domain-skill-pack

OpenClaw-specific search skills providing multi-source parallel search, content extraction, and document parsing. Three composable skills: search-layer, content-extract, mineru-extract.

Structure

- Four-source parallel search (Brave + Exa + Tavily + Grok) - Intent-aware retrieval (factual/status/comparison/tutorial/exploratory/news/resource) - Chain reference tracing (GitHub issues, HN, Reddit threads) - Research-light lane for deep queries

search-layer/ -- SKILL.md + scripts + references
content-extract/ -- URL to clean Markdown with anti-scraping fallback
mineru-extract/ -- MinerU API wrapper for PDF/Office/HTML to Markdown

Extractable Value

Processes

**Deep research methodology** (specializations/shared/): The two-path retrieval architecture (standard retrieval path + thread-pulling path) combined with intent classification is a solid research methodology. Maps to a babysitter process: classify intent -> multi-source retrieve -> score/dedup -> optionally deep-pull threads -> synthesize.
**Content extraction pipeline** (specializations/shared/): The graduated fallback pattern (direct fetch -> anti-scraping fallback -> MinerU parse) for converting arbitrary URLs to clean markdown is a reusable extraction process.

Plugin Ideas

**Multi-source search plugin:** A babysitter marketplace plugin providing a search task kind that dispatches to multiple search APIs in parallel (Brave/Exa/Tavily), with intent-aware routing and result fusion. Install.md configures API keys per source.

Library Mapping

Extractable Process	Library Status	Action	Existing Path	Target Placement
Deep Research Methodology	NEW	Two-path retrieval architecture with intent classification and multi-source search	-	specializations/shared/deep-research-methodology.js
Content Extraction Pipeline	NEW	Graduated fallback pattern (direct fetch → anti-scraping → MinerU parse) for URL to Markdown	-	specializations/shared/content-extraction-pipeline.js
Multi-Source Parallel Search	NEW	Four-source parallel search with result fusion and deduplication	-	specializations/shared/multi-source-parallel-search.js
Intent-Aware Retrieval	NEW	Search intent classification (factual/status/comparison/tutorial/exploratory/news/resource)	-	specializations/shared/intent-aware-retrieval.js

Plugin Marketplace Mapping

Plugin Idea	Marketplace Status	Action	Existing Plugin	Target Placement
Multi-Source Search Engine	NEW	Search task kind dispatching to multiple APIs (Brave/Exa/Tavily) with API key configuration	-	plugins/a5c/marketplace/plugins/multi-source-search-engine/

SKIP

OpenClaw-specific skill loading mechanics
The Grok integration (model-specific, not generalizable)