ma-fulltext-management - Full-text collection and management
Collect and manage full-text PDFs, track provenance, and prepare documents for data extraction
Tags
Updated: 2026-03-23Capabilities
Typical Inputs
Typical Outputs
What this skill does
- create directory structure
- build manifest file
- read screening records
- write manifest CSV
- run web extraction
- query PubMed database
- query ClinicalTrials.gov
- update confidence scores
- flag low-confidence studies
- query Unpaywall API
- download PDF files
- render PDF previews
- run AI screening
- compute inter-rater agreement
- resolve screening conflicts
Inputs
- 03_screening/round-01/included.bib
- UNPAYWALL_EMAIL environment variable
Outputs
- 04_fulltext/manifest.csv
- 04_fulltext/unpaywall_results.csv
- 04_fulltext/fulltext_decisions.csv
- 04_fulltext/ft_agreement.md
- 04_fulltext/README.md
- 04_fulltext/PDF files
- 04_fulltext/previews
Requirements
- UNPAYWALL_EMAIL in .env
- pdftoppm or mutool installed