literature-review

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

• do a literature review
• search the literature on
• find papers about
• systematic review of
• meta-analysis search

What it does

literature-review is a Claude Code skill from K-Dense AI’s scientific-agent-skills repo. It turns Claude into a systematic reviewer that queries multiple academic databases in parallel — PubMed, arXiv, bioRxiv, Semantic Scholar, and others — then synthesizes results into a professionally formatted markdown document or PDF with verified citations in the target style (APA, Nature, Vancouver, etc.).

A session produces a structured literature review document: an introduction framing the search scope, a methods section documenting search strings and inclusion/exclusion criteria, a synthesized findings narrative, and a formatted reference list. Citations are verified, not hallucinated.

When to use it

Reach for it when:

• You’re starting a research project and need to map the existing evidence base before writing
• You’re writing the Introduction or Background section of a manuscript and need a sourced synthesis rather than a summary from memory
• You need a documented, reproducible search for a systematic review or meta-analysis

When not to reach for it:

• You already know the specific papers you need — use paper-lookup for direct DOI/PMID retrieval
• You only need a quick answer to a factual question — research-lookup or parallel-web is faster

Install

Copy the SKILL.md from K-Dense AI’s literature-review folder into .claude/skills/literature-review/ in your project.

Trigger phrases: “do a literature review”, “search the literature on”, “systematic review of”.

What a session looks like

A typical session has three phases:

1. Search design. Claude asks for the topic, PICO elements if applicable, and date range. It proposes search strings across databases and you approve or adjust.
2. Multi-database retrieval. Queries run in parallel; results are de-duplicated, filtered against inclusion criteria, and ranked by relevance.
3. Synthesis document. Claude drafts the review narrative, groups findings by theme or study type, and appends a formatted reference list. You receive a markdown file ready for further editing.

Receipts

Where it works well:

• Established biomedical topics with strong PubMed coverage — retrieval is comprehensive and citation verification is reliable
• Generating the documented search methodology section required by systematic review journals

Where it backfires:

• Very recent preprints on fast-moving topics — database latency means the last 2–4 weeks of arXiv/bioRxiv may be incomplete
• Niche sub-fields where the relevant literature is in conference proceedings not indexed in the supported databases

Pattern that works: run the search first with broad terms, review the candidate list, then narrow with MeSH or field-specific vocabulary for the final synthesis pass.

Source and attribution

Originally authored by K-Dense Inc.. The canonical SKILL.md lives in the literature-review folder of their public 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.