dac-submission

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes standard command-line utilities including pdfinfo, pdftotext, and grep to inspect local PDF files for sensitive metadata and keywords. These operations are performed locally and are intended for the skill's primary purpose of anonymizing research papers.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided PDF files (paper.pdf) to detect sensitive information. Ingestion Point: paper.pdf. Boundary Markers: None. Capability Inventory: pdfinfo, pdftotext, grep. Sanitization: Read-only inspection; extracted content is searched for specific patterns but is not interpreted or executed as agent instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 12:05 PM