usenixsec-camera-ready
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a bash script that executes local system utilities (
pdftotext,pdffonts,grep,awk) to verify formatting, font embedding, and metadata of the final paper. - [EXTERNAL_DOWNLOADS]: The skill uses
curlto verify the availability of URLs extracted from the paper. It performs HTTP HEAD requests to external servers to confirm that artifact mirrors are live, though it does not execute or process the resulting response bodies. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes text extracted from a user-provided PDF file (
final.pdf). This content enters the agent's context for verification purposes, creating a potential surface where instructions embedded in the document could attempt to influence the agent's behavior.
Audit Metadata