large-document-processing

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process large, untrusted documents (PDF, DOCX, EPUB) for AI training and RAG systems.
  • Ingestion points: The skill reads external files using fitz.open() (PDF), python-docx (DOCX), and NWTEpubParser (EPUB), as described in SKILL.md.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded instructions are visible in the processing logic to prevent the agent from obeying prompts found within document content.
  • Capability inventory: The skill manages file system operations, including reading source documents and writing processed outputs to the output/processed_document/ directory.
  • Sanitization: The provided text chunking implementation does not include sanitization or filtering logic to neutralize potential instructions contained within the extracted text.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 06:45 AM
Security Audit — agent-trust-hub — large-document-processing