pymupdf
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the use of PyMuPDF and PyMuPDF4LLM for parsing external documents (PDF, Office, etc.), creating a potential surface for indirect prompt injection if malicious instructions are embedded in the documents being processed.
- Ingestion points: Data enters the system via document parsing methods described in
references/api.mdandreferences/tutorials.md. - Boundary markers: The instructions do not specify any markers or rules to isolate parsed content from agent instructions.
- Capability inventory: The described tools include capabilities for file creation (
Document.save()), image processing, and extraction of structured data. - Sanitization: There are no documented steps for sanitizing extracted content before it is processed by the AI agent.
Audit Metadata