← ClaudeAtlas

tyler_medlisted

Convert a folder of MEDICAL/clinical academic PDFs into a token-efficient markdown wiki + evidence table for literature review. Use when the user has a folder of clinical journal PDFs (JVS, JAMA, NEJM, Ann Surg, Cochrane, etc.) to process into .md files with study-design, sample-size, data-source, and DOI metadata. Trigger phrases include "build my medical wiki", "process my clinical papers", "convert these journal PDFs", "/tyler_med".
rlarson94170/tyler_med · ★ 0 · Data & Documents · score 55
Install: claude install-skill rlarson94170/tyler_med
# PDF-to-Wiki Skill — Medical Edition ## Purpose Converts a directory of clinical/medical academic PDFs into a structured, two-tier markdown wiki optimised for token efficiency AND for literature-review workflows. It is a medical-tuned fork of the `tyler` skill: it drops the economics machinery (JEL codes, JSTOR filters) and adds clinical metadata. A typical 12-page clinical paper costs ~12,000 tokens to read in full. This skill produces: - **Tier 1 — Index** (`index.md`): ~450 tokens/paper, **grouped by study design** in evidence-hierarchy order (guidelines → systematic reviews → RCTs → cohort → … ). Each entry has title, authors, year, journal, DOI link, sample size (N), data source, and abstract. - **Tier 1b — Evidence table** (`index.csv` + `index.json`): the same metadata as a spreadsheet-ready table — a ready skeleton for a PRISMA-style evidence table. - **Tier 2 — Full papers** (`papers/*.md`): cleaned markdown with rich YAML frontmatter. Read on demand for detail. - **References** (`references/*.md`): trimmed reference lists kept separately, so the main files stay lean but citations remain available for snowball searching. ## What makes it medical-tuned - **Reliable titles**: reads the PDF's embedded metadata + DOI (via PyMuPDF), so it recovers correct titles even when a filename is truncated/abbreviated. A curated `YEAR JOURNAL - Title` filename still takes precedence when present. - **DOI extraction** → frontmatter + clickable `https://doi.org/