image-ad-clone
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze and reverse-engineer existing ad images. This creates a surface for indirect prompt injection where an attacker could embed malicious instructions within the image text or layout that the LLM might interpret as commands during the visual analysis phase.
- Ingestion points: User-provided ad images (processed in Phase 2).
- Boundary markers: None identified in the workflow instructions to separate image-derived content from system instructions.
- Capability inventory: Executes Python scripts (
generate_image.py) and writes/appends to documentation files (prompt-library.md). - Sanitization: No sanitization or filtering of text extracted from images is mentioned before processing.
- [COMMAND_EXECUTION]: The workflow involves executing local Python scripts to validate generated prompts. While these are part of the skill's own environment, they represent a mechanism for local code execution.
- Evidence: References to
skills/chatgpt-image-ad/scripts/generate_image.pyandskills/nano-banana-image-ad/scripts/generate_image.pybeing called to generate images for validation.
Audit Metadata