brand-voice-ai-training
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill content is entirely instructional, focusing on brand voice analysis and prompt template generation without any malicious intent or harmful commands. Evaluation of indirect prompt injection surfaces (Category 8) identifies that although the skill ingests untrusted sample content in Step 1, it lacks any executable capabilities, tool permissions, or network access to manifest a threat.
- [NO_CODE]: No code snippets, scripts, or tool invocations are present in the skill, precluding any risks related to command execution, privilege escalation, or remote code vulnerabilities.
- [PROMPT_INJECTION]: No patterns of system prompt extraction, safety bypass, or instruction overrides were detected. Role-playing instructions are used legitimately for voice training and alignment.
- [DATA_EXFILTRATION]: While the process involves gathering brand-specific content samples from the user, the skill does not include any tools, network operations, or instructions to transmit this information to external destinations.
- [EXTERNAL_DOWNLOADS]: No external packages or remote scripts are requested or executed. References to methodology (e.g., Evelyn 2025, Mizrahi 2024) are used for citation and are not hyperlinked to remote code.
Audit Metadata