fuite-de-surface
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely focused on output formatting and persona simulation. It does not contain any code, scripts, or instructions that interact with the underlying system.
- [PROMPT_INJECTION]: Although the skill modifies the agent's behavior by adding a narrative layer, it does not attempt to bypass safety filters, extract system prompts, or disregard previous instructions. It specifically limits its own behavior in sensitive contexts like medicine, law, or finance.
- [DATA_EXFILTRATION]: No network tools, URL references, or sensitive file paths (such as credentials or SSH keys) are present in the instructions or references.
- [COMMAND_EXECUTION]: There are no shell commands, subprocess calls, or execution patterns within the files.
Audit Metadata