6verity
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements standard document validation workflows using local scripts and widely-used formatting tools like Typst, LaTeX, and 7-Zip. No malicious patterns such as exfiltration or unauthorized command execution were identified.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided paper source files and result data, which is necessary for its verification task.\n
- Ingestion points: The skill reads LaTeX/Typst files (main.tex, main.typ) and project reports (all_results.json).\n
- Boundary markers: There are no explicit instructions for the agent to disregard instructions potentially embedded within the documents it validates.\n
- Capability inventory: The skill uses Bash for compilation and Read/Write tools for report generation.\n
- Sanitization: The check script uses regex-based scanning for placeholders and keywords, which does not involve executing the file content as code.
Audit Metadata