gpt-taste
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses directive language and role-play instructions to define the agent's behavior as a design engineer. These commands are intended to improve output quality and stylistic consistency rather than bypass security protocols.
- [INDIRECT_PROMPT_INJECTION]: The skill uses the length of the user's prompt as a seed for randomized design selections. While this utilizes external input to influence behavior, the resulting actions are limited to selecting visual layouts and do not introduce execution or exfiltration risks.
- [EXTERNAL_DOWNLOADS]: The skill refers to picsum.photos for generating placeholder imagery. This is a well-known, safe service for developers and does not represent a malicious external dependency.
- [DYNAMIC_EXECUTION]: The instructions mandate a simulated reasoning step described as a "mock Python output." This is a prompt-engineering technique to ensure variety in the agent's internal logic and does not involve the actual execution of Python code on the system.
Audit Metadata