Literature Review

Literature Review is a skill in K-Dense's Claude Scientific Writer (MIT) for systematic reviews, scoping reviews, and the literature section of a paper or thesis. Version 1.5 searches several sources, screens and synthesizes by theme, verifies citations, and writes Markdown and PDF output.

Key Features

• Seven phases: planning and scoping, systematic search, title/abstract then full-text screening (with counts kept for a PRISMA flow), data extraction and quality appraisal, synthesis, citation verification, and document generation.
• Search stack: the primary tool is parallel-cli search from the Parallel web skill, with --include-domains to focus on scholarly sites and parallel-cli extract to read promising pages. Database skills such as gget (PubMed, bioRxiv) and bioservices cover specialist sources. The skill asks for at least three databases.
• Bundled scripts: verify_citations.py checks DOIs and formats citations, search_databases.py deduplicates and formats results, and generate_pdf.py turns Markdown into PDF through Pandoc and LaTeX.
• Citation styles: APA, Nature, Vancouver, Chicago, and IEEE guidance in references/citation_styles.md, plus a full review template in assets/review_template.md.
• Figures by default: the skill still instructs the agent to include one or two generated figures, such as a PRISMA flow diagram, via Scientific Schematics.

Use Cases

• A systematic or scoping review where every search string and date must be reproducible.
• The related-work chapter of a thesis, organized by theme rather than paper by paper.
• A quick state-of-the-art survey to find research gaps before a grant.

Pricing and Access

The skill is free (MIT). Parallel search needs a Parallel account (parallel-cli login or PARALLEL_API_KEY), and generated figures use an optional OPENROUTER_API_KEY; both bill through those services. Pandoc and a LaTeX distribution are needed for PDF output.

Getting Started

1. Install the plugin in Claude Code: /plugin marketplace add https://github.com/K-Dense-AI/claude-scientific-writer, then /plugin install claude-scientific-writer.
2. Install the search CLI: uv tool install "parallel-web-tools[cli]", then parallel-cli auth.
3. Run python scripts/generate_pdf.py --check-deps, then ask for a review with clear inclusion and exclusion criteria.

Limitation: search coverage is only as good as the queries and databases used, and paywalled full text may be unavailable. For a formal systematic review, dual human screening is still recommended by the skill itself.

Frequently Asked Questions

How is it different from Research Lookup?

Research Lookup compiles a verified evidence packet for a manuscript. Literature Review adds protocol, screening, exclusion reasons, and quality appraisal for a PRISMA-style review.

Does it check that references are real?

It verifies DOIs with a script, but the skill still expects a human to confirm each source supports the claim.

Alternatives

• NotebookLM: Google's notebook for asking questions over sources you upload.
• Perplexity AI: fast cited web answers, without a review protocol.
• Citation Management Skill: BibTeX extraction and validation to pair with this skill.

Conclusion

Literature Review suits researchers who need a documented, repeatable review process inside Claude Code. It depends on paid search and optional image APIs, and it does not replace human screening. See more research tools in the skills hub.