Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill provides instructions to run
sudo apt-get installfor system dependencies, which requires administrative access to the environment. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external PDF files which could contain adversarial prompts.
- Ingestion points: Data from untrusted PDF files is loaded into the agent context via extraction and rendering tools as specified in SKILL.md.
- Boundary markers: There are no instructions to use delimiters or ignore content within the ingested PDF data.
- Capability inventory: The skill can execute shell commands for processing, write output files, and install new software.
- Sanitization: No validation or filtering of the PDF content is described in the provided workflow.
- [COMMAND_EXECUTION]: The skill involves executing shell commands (
pdftoppm,pip,brew,apt-get) where input is provided via variables like$INPUT_PDF, potentially allowing for command injection if input is untrusted. - [EXTERNAL_DOWNLOADS]: The skill installs several Python packages (
reportlab,pdfplumber,pypdf) and system tools from public repositories during the environment setup phase.
Audit Metadata