visa-doc-translate
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands such as 'sips' for image conversion and uses 'pip' or 'brew' to install necessary dependencies.
- [REMOTE_CODE_EXECUTION]: The skill generates a Python script at runtime to create a bilingual PDF and executes it, which involves running dynamically created code.
- [PROMPT_INJECTION]: The instructions require the agent to operate 'WITHOUT asking for confirmation', removing user oversight. This increases the risk from indirect prompt injection, as the skill extracts and processes untrusted text from user-uploaded images via OCR. Evidence: Ingestion point (user-provided images via OCR extraction), Boundary markers (absent), Capabilities (shell command execution and dynamic Python script execution), Sanitization (none provided for the extracted text before processing).
Audit Metadata