tyler_med v0.1.0

First tagged release of tyler_med — a Claude Code skill that turns a folder of clinical-journal PDFs into a token-efficient Markdown wiki and evidence table for literature review. Medical-tuned fork of the tyler skill from econtools (MIT).

Highlights

• Two-tier wiki: index.md grouped by study design (evidence hierarchy) + full papers/*.md with rich YAML frontmatter; read the index cheaply, open a paper only on demand.
• Evidence table: index.csv / index.json — a spreadsheet-ready, PRISMA-style skeleton.
• Reliable metadata: titles from embedded PDF metadata + filename (with near-duplicate ligature-gap repair), DOI extraction with clickable links, and mojibake/ligature cleanup (n ¼ 4,894 → n = 4,894).
• Clinical classification: study design (RCT, systematic review, meta-analysis, cohort, guideline/consensus, …), sample size, data source (NSQIP, VQI, Medicare/CMS, Cochrane, …), and trial/PROSPERO registration IDs.
• Duplicate detection by shared DOI or year+title.
• Publisher-aware abstract extraction, including the Elsevier two-column and BMJ Open layouts.

Robustness

• Resumable runs: state saved after every file; index rebuilt from state (--index-only); --time-budget for time-capped, no-background sandboxes (re-run until ALL_DONE).
• OCR control (--ocr auto|off|force): auto (default) skips OCR for born-digital journal PDFs — ~4–5× faster per file with identical extraction — and OCRs only true scans.

Requirements

• Python 3.9+ and pymupdf4llm (bundles PyMuPDF / fitz).

Licensed under MIT. See README.md and SKILL.md for usage.