pdf-to-markdown

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The utility script bin/pdf-to-markdown fetches version metadata and binary archives from the vendor's distribution network.
  • Evidence: The script references https://agent-cdn.nutrient.io/pdf-to-markdown/LATEST to check for updates and https://agent-cdn.nutrient.io/pdf-to-markdown/{release_id}/{target_id}.tar.gz for downloads.
  • [REMOTE_CODE_EXECUTION]: The skill downloads an executable binary at runtime, installs it to the user's local directory, and executes it to process documents.
  • Evidence: The install_archive function in bin/pdf-to-markdown downloads, extracts, and places the binary in ~/.local/share/nutrient/cli/, which is then invoked via exec at the end of the script.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external PDF files, which represents a potential surface for indirect prompt injection if the PDF content contains malicious instructions.
  • Ingestion points: Accepts PDF files or directories through the INPUT.pdf and INPUT_DIR/ arguments.
  • Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are configured for the extracted Markdown output.
  • Capability inventory: The skill executes the conversion binary and utilizes shell commands like grep and head for content inspection.
  • Sanitization: The skill performs extraction of text/tables but does not explicitly sanitize the resulting Markdown for downstream agent instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 03:23 PM