multi-image-consistency
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No attempts to override system instructions or safety filters were detected. The skill uses standard instructional language to guide the AI's behavior.
- [DATA_EXFILTRATION]: No network operations, hardcoded credentials, or sensitive file path access were identified. The skill operates entirely on provided text and outputs generated prompts.
- [REMOTE_CODE_EXECUTION]: No remote code patterns, script downloads, or package installations were found.
- [COMMAND_EXECUTION]: No shell commands, subprocess spawning, or dynamic execution patterns are present.
- [OBFUSCATION]: No Base64, hex encoding, zero-width characters, or hidden text patterns were detected.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied planning documents (PRDs, scripts, etc.), which represents a potential injection surface. However, the risk is minimal as the output is strictly text-based prompts for image generation, and the instructions focus on structured extraction of visual facts.
Audit Metadata