seed-pdf-import

Pass

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it is designed to ingest and process untrusted PDF content.
  • Ingestion points: External PDF files are read using vision capabilities or extraction libraries as described in SKILL.md and references/pdf-extraction.md.
  • Boundary markers: There are no instructions or delimiters provided to distinguish between document content and potential malicious instructions embedded within the PDF text or metadata.
  • Capability inventory: The skill uses various subprocess calls in SKILL.md and references/pdf-extraction.md, including pdfimages, pdftoppm, pdftotext, and the seed-cli for document creation and publishing. It also writes files to the /tmp/ directory.
  • Sanitization: The skill lacks evidence of sanitizing, validating, or escaping extracted content before it is processed by the agent or passed as arguments (e.g., --name, --display-author) to CLI tools.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of local system utilities and vendor-provided CLI tools to handle file conversion and data publishing.
  • The workflow involves running shell commands such as pdfimages, pdftoppm, and seed-cli (from the @seed-hypermedia/cli package) with parameters derived from the extracted PDF data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 22, 2026, 11:16 AM