Embed badge
Show
Style
[](https://versuz.dev/skills/vivekkarmarkar-claude-code-os-skills-litreview)Free SKILL.md scraped from GitHub. Clone the repo or copy the file directly into your Claude Code skills directory.
npx versuz@latest install vivekkarmarkar-claude-code-os-skills-litreviewgit clone https://github.com/VivekKarmarkar/claude-code-os.gitcp claude-code-os/SKILL.MD ~/.claude/skills/vivekkarmarkar-claude-code-os-skills-litreview/SKILL.md# Literature Review — Research, Organize, Present Conduct a thorough academic literature review: find papers, map connections, read and summarize them, organize everything locally, and produce a polished presentation. ## Arguments `<research topic or question>` — What the literature review is about. Examples: - `/litreview CRISPR delivery mechanisms for in vivo gene therapy` - `/litreview transformer architectures for time series forecasting` - `/litreview effects of microplastics on marine biodiversity` If no arguments, ask the user for their research topic, scope, and any constraints (date range, specific journals, key authors to include). ## Pre-flight Check 1. Call `mcp__claude-in-chrome__tabs_context_mcp` to verify Chrome MCP is connected. If not, stop and tell the user to connect it — Chrome is needed for Connected Papers and Napkin AI. 2. Confirm the user's home directory is accessible for saving files. --- ## Phase 1: Scope the Review Before searching, clarify with the user: 1. **Topic focus** — What specific question or area? 2. **Depth** — Quick survey (5-10 papers) or deep dive (20-30 papers)? 3. **Date range** — Recent only (last 5 years)? Or historical too? 4. **Key papers** — Any seed papers the user already knows about? 5. **Audience** — Who is the presentation for? (determines technical depth) Use answers to guide search strategy. --- ## Phase 2: Search for Papers Use multiple sources in parallel to build a comprehensive paper list. ### Source 1: Web Search (Google Scholar) Use `WebSearch` with targeted queries: - `"<topic>" site:scholar.google.com` - `"<topic>" review OR survey OR meta-analysis` - `"<topic>" <key author names>` - Vary queries to cover different angles of the topic. Extract: title, authors, year, citation count, DOI/URL. ### Source 2: PubMed (for biomedical/life sciences topics) If the topic is biomedical, use the PubMed MCP tools: - `mcp__claude_ai_PubMed__search_articles` with relevant queries - `mcp__claude_ai_PubMed__get_article_metadata` for detailed info - `mcp__claude_ai_PubMed__find_related_articles` to expand from key papers - `mcp__claude_ai_PubMed__get_full_text_article` for open-access full text ### Source 3: Connected Papers (Citation Graph) For the top 3-5 most relevant papers found so far: 1. Open a new Chrome tab: `mcp__claude-in-chrome__tabs_create_mcp` 2. Navigate to `https://www.connectedpapers.com/` 3. Search for the paper by title or DOI 4. Once the graph loads, take a snapshot to read the connected papers 5. Identify: - **Prior work** — foundational papers this paper builds on - **Derivative work** — papers that cite this one - **Clusters** — groups of related papers that form sub-topics 6. Add newly discovered important papers to the list ### Source 4: arXiv (for CS, physics, math topics) If the topic is in CS/physics/math, use `WebSearch` with: - `site:arxiv.org "<topic>"` - Check for recent preprints not yet in other databases ### Compile the Master List Create a deduplicated list of all found papers, ranked by: 1. Relevance to the research question 2. Citation count (influence) 3. Recency 4. Whether it's a review/survey paper (prioritize these) Present the list to the user: ``` Found N papers. Top candidates: # Year Citations Title Authors 1 2024 342 <title> <first author et al.> 2 2023 891 <title> <first author et al.> ... ``` Ask: "Which papers should I include? All of them, or a subset?" --- ## Phase 3: Read and Analyze Papers For each selected paper: ### 3a. Get the Full Text Try in this order: 1. PubMed Central full text (`mcp__claude_ai_PubMed__get_full_text_article`) if available 2. Open-access PDF via DOI link (use Chrome to navigate and check) 3. arXiv PDF if available 4. Fall back to abstract only if full text isn't freely available ### 3b. Read and Extract For each paper, extract: - **Core claim / thesis** — What is this paper arguing or demonstrating? - **Methods** — What approach did they use? - **Key findings** — What are the main results? - **Limitations** — What do the authors acknowledge as limitations? - **Relevance** — How does this connect to the research question? - **Key figures/tables** — Note which figures are most important - **Key quotes** — 1-2 short quotes that capture the paper's essence ### 3c. Map Relationships As you read, build a mental model of how papers relate: - Which papers agree/disagree? - What's the chronological progression of ideas? - Are there competing schools of thought? - What gaps exist in the literature? --- ## Phase 4: Create Folder Structure Create an organized directory structure on the user's computer: ``` ~/lit-reviews/<topic-slug>/ ├── papers/ │ ├── AuthorYear_ShortTitle.pdf │ ├── AuthorYear_ShortTitle.pdf │ └── ... ├── notes/ │ ├── AuthorYear_ShortTitle.md │ └── ... ├── presentation/ │ └── lit-review-<topic-slug>.pptx ├── summary.md └── bibliography.md ``` ### Naming Convention - **Folder name:** lowercase, hyphen-separated topic slug (e.g., `crispr-delivery-mechanisms`) - **Paper PDFs:** `LastnameYear_TwoOrThreeWordTitle.pdf` (e.g., `Chen2024_CRISPRDelivery.pdf`) - **Note files:** Same as PDF but `.md` extension - **No spaces in filenames.** Use CamelCase or hyphens. ### Download Papers For each paper with a freely available PDF: 1. Navigate to the PDF URL in Chrome 2. Ask the user for permission to download 3. Save to the `papers/` directory with proper naming **Important:** Only download open-access papers. Do not attempt to bypass paywalls. If a paper is behind a paywall, note it in `bibliography.md` with access instructions. ### Create Note Files For each paper, create a markdown note in `notes/`: ```markdown # <Full Paper Title> **Authors:** <full author list> **Year:** <year> **Journal:** <journal name> **DOI:** <doi> **Citations:** <count> ## Summary <2-3 paragraph summary> ## Key Findings - <finding 1> - <finding 2> - <finding 3> ## Methods <brief methods description> ## Relevance to Review <how this connects to the research question> ## Limitations <noted limitations> ## Key Quotes > "<quote 1>" > "<quote 2>" ``` ### Create bibliography.md A complete bibliography in a consistent citation format (APA by default): ```markdown # Bibliography 1. Author, A. B., & Author, C. D. (Year). Title. *Journal*, volume(issue), pages. https://doi.org/xxx 2. ... ``` ### Create summary.md A narrative literature review document: ```markdown # Literature Review: <Topic> **Date:** <today> **Papers reviewed:** N ## Overview <1-2 paragraph overview of the field> ## Themes ### <Theme 1> <discussion with citations> ### <Theme 2> <discussion with citations> ## Gaps in the Literature <what's missing, what needs more research> ## Conclusions <synthesis of findings> ## Bibliography <reference to bibliography.md> ``` --- ## Phase 5: Create Presentation Use the `/pptx` skill to create a literature review presentation. The presentation should follow this structure: 1. **Title slide** — Topic, your name (ask user), date 2. **Research Question** — What this review investigates 3. **Search Methodology** — Where you searched, how many papers found/selected 4. **Timeline / Field Overview** — How the field has evolved 5. **Theme slides** (2-4 slides per theme) — Key findings organized by theme, with citations 6. **Comparison / Debate** — Where papers agree/disagree 7. **Gaps & Future Directions** — What's missing in the literature 8. **Conclusions** — Key takeaways 9. **Bibliography** — Full reference list (can be multiple slides) Save to `presentation/lit-review-<topic-slug>.pptx`. --- ## Phase 6: Polish with Napkin AI Use Chrome to polish the presentation slides: 1. Open a new Chrome tab 2. Navigate to `https://app.napkin.ai/page/create` 3. For each key content slide (skip title and bibliography): - Paste the slide's text content into Napkin AI - Let Napkin generate visual representations (diagrams, infographics) - Take a screenshot of the generated visual - Save the visual to `presentation/napkin-visuals/` 4. Report to the user which slides have Napkin visuals available, so they can manually insert their favorites into the PPTX. **Note:** Napkin AI generates visuals from text — it works best for conceptual content, comparisons, timelines, and process flows. Not every slide will benefit. --- ## Final Report ``` Literature Review Complete ═══════════════════════════════════════════ Topic: <topic> Papers found: N (searched) Papers reviewed: N (selected) Papers downloaded: N (open access) Created: ~/lit-reviews/<topic-slug>/ ├── papers/ N PDFs ├── notes/ N markdown summaries ├── presentation/ 1 PPTX + N Napkin visuals ├── summary.md Narrative review └── bibliography.md Full reference list Presentation: <path to pptx> ═══════════════════════════════════════════ ``` --- ## Rules 1. **Only download open-access papers.** Never bypass paywalls, use sci-hub, or access papers illegally. 2. **Ask before every download.** Each PDF download needs explicit user permission. 3. **Cite everything.** Every claim in the summary and presentation must reference a specific paper. 4. **Be honest about access.** If you can only read an abstract, say so — don't pretend you read the full paper. 5. **PubMed attribution.** When using PubMed data, always cite PubMed and include DOIs as required by the MCP tool. 6. **Respect copyright.** Never reproduce large portions of papers. Summarize in your own words, use short quotes only. 7. **Let the user steer.** Present paper lists and theme choices for approval — don't assume what's relevant. 8. **Save as you go.** Don't wait until the end to create files. Build the folder structure early and populate incrementally.