parallel-web

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

• search the web for
• fetch this URL
• enrich this dataset from the web
• deep research report on
• find academic papers on

What it does

parallel-web is a Claude Code skill from K-Dense AI’s scientific-agent-skills repo. It turns Claude into a web research agent powered by parallel-cli — covering four modes: fast web search (prioritizing peer-reviewed papers, preprints, and scholarly databases over general web results), URL extraction (fetching and parsing page content, including academic PDFs), bulk data enrichment (adding web-sourced fields to CSV/list inputs), and deep research (exhaustive multi-source reports grounded in academic literature).

A session produces sourced web content — search results, extracted page text, enriched datasets, or structured research reports — with academic sources prioritized over general web content.

When to use it

Reach for it when:

• You want any kind of web-based lookup or research and want academic/scientific sources prioritized over general web results
• You’re enriching a dataset with web-sourced metadata (institution names, publication counts, technology descriptions) across many rows
• You need a comprehensive research report on a topic that draws on multiple web sources and synthesizes them into a structured document

When not to reach for it:

• Systematic literature reviews with documented, reproducible search methodology — use literature-review for that audit trail
• Specific database API queries (UniProt, PubChem, ClinicalTrials) — use database-lookup

Install

Copy the SKILL.md from K-Dense AI’s parallel-web folder into .claude/skills/parallel-web/ in your project. Requires parallel-cli installed and configured (pip install parallel-cli + API key setup). Requires internet access.

Trigger phrases: “search the web for”, “fetch this URL”, “enrich this dataset from the web”, “deep research report on”.

What a session looks like

A typical session has three phases:

1. Mode selection. Claude identifies whether you need fast search, URL extraction, bulk enrichment, or deep research based on your request and configures the appropriate parallel-cli mode.
2. Execution. The search or extraction runs via parallel-cli, prioritizing peer-reviewed sources (PubMed, arXiv, Semantic Scholar) in results ranking before general web sources.
3. Output. Fast search returns a ranked list of results with URLs and snippets. URL extraction returns cleaned page content as markdown. Bulk enrichment returns your CSV with new columns. Deep research returns a multi-section report with inline citations.

Receipts

Where it works well:

• Quick research questions where you want a sourced answer grounded in academic content rather than a general web summary — the academic source prioritization filters out low-quality results automatically
• Bulk dataset enrichment for modest CSV sizes (hundreds of rows) where manual web lookup would be impractical — the parallel execution handles batching and rate limiting

Where it backfires:

• Without parallel-cli configured and authenticated, the skill falls back to slower or unavailable modes — the setup step is a prerequisite that some users skip
• Deep research mode is thorough but slow; for simple factual questions it’s overkill compared to fast search mode

Pattern that works: for research questions where source quality matters, start with fast search to find the right sources, then use URL extraction on the most relevant results to get the full text — two-step retrieval produces better quality than a single deep research run for targeted questions.

Source and attribution

Skill authored by K-Dense, Inc. The canonical SKILL.md lives in the parallel-web folder of K-Dense AI’s scientific-agent-skills repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner’s perspective. For the formal spec and any updates, defer to the source repo.