pdf-to-text

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The bin/pdf-to-text script fetches platform-specific binaries and versioning metadata from https://agent-cdn.nutrient.io. These resources belong to the vendor's official infrastructure and are used to install the underlying extraction tool.
  • [REMOTE_CODE_EXECUTION]: The skill downloads a compressed archive from the vendor's CDN, extracts it, and executes the resulting binary. This behavior is part of the tool's intended installation and update process from its developer.
  • [COMMAND_EXECUTION]: The shell wrapper invokes the extracted utility via exec to perform PDF-to-text conversion tasks.
  • [DYNAMIC_EXECUTION]: The script performs runtime extraction and execution of an external tool as a standard part of its operation.
  • [INDIRECT_PROMPT_INJECTION]: The skill extracts text from untrusted PDF documents, which can lead to the agent processing hidden instructions. 1. Ingestion points: PDF files passed to the conversion command. 2. Boundary markers: None identified in the wrapper or instructions. 3. Capability inventory: Local file writing and execution of the converter binary. 4. Sanitization: No sanitization of extracted text is performed.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 04:16 PM
Security Audit — agent-trust-hub — pdf-to-text