Skills
Navigate
skills/brunoasm/my_claude_skills/extract-from-pdfs/Gen Agent Trust Hub

extract-from-pdfs

Pass

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted scientific literature (PDFs, BibTeX, and RIS files), creating a surface for indirect prompt injection.
  • Ingestion points: External data is ingested through scripts/01_organize_metadata.py and scripts/03_extract_from_pdfs.py.
  • Boundary markers: The skill mitigates risks by using XML delimiters (e.g., <title>, <abstract>, <analysis>, <output>) in its prompts within scripts/02_filter_abstracts.py and scripts/03_extract_from_pdfs.py to separate untrusted data from instructions.
  • Capability inventory: The pipeline has file-writing capabilities across multiple formats (CSV, JSON, Python pickle, R RDS, SQLite) via scripts/06_export_database.py.
  • Sanitization: The skill uses json-repair and jsonschema in scripts/04_repair_json.py to validate and sanitize the structured data extracted by the model.
  • [COMMAND_EXECUTION]: The script scripts/06_export_database.py dynamically generates auxiliary Python (.py), R (.R), and SQL (.sql) scripts to assist users with data loading. These scripts are generated from predefined templates using local file paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 11, 2026, 03:38 AM
Security Audit — agent-trust-hub — extract-from-pdfs