pdfcraft
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF documents through operations like OCR and text extraction, creating an attack surface for indirect prompt injection where instructions embedded in documents could influence agent behavior.
- Ingestion points: External files are ingested and processed via tools like
file.openand thepdfcraft-cli batchcommand. - Boundary markers: The skill instructions do not provide specific delimiters or warnings to help the agent distinguish between tool instructions and document content.
- Capability inventory: The agent has the capability to read and write files within configured root directories and perform document modifications.
- Sanitization: Content extracted from PDF files is provided to the agent without evidence of sanitization or filtering.- [EXTERNAL_DOWNLOADS]: The skill references documentation and configuration resources hosted on external GitHub repositories.
- Evidence: References point to documentation located within the
storytold/pdfcraftrepository on GitHub for automation conventions and batch processing.
Audit Metadata