code-from-image
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (LOW): The skill's primary workflow involves the agent implementing and running logic extracted from images. This creates a risk if the logic is attacker-controlled.
- [EXTERNAL_DOWNLOADS] (LOW): Recommends the dynamic installation of Python packages 'pillow' and 'pytesseract'. While these are standard libraries, manual installation via agent commands bypasses static environment controls.
- [PROMPT_INJECTION] (LOW): Vulnerable to indirect prompt injection. 1. Ingestion points: PNG and JPG image files processed via OCR tools. 2. Boundary markers: Absent; there are no instructions to wrap or isolate the extracted code. 3. Capability inventory: The agent is tasked with implementing and executing the resulting code, which could include file or network access. 4. Sanitization: Absent; the skill does not suggest safety checks or sandboxing for the extracted logic.
Audit Metadata