voice-builder
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions designed to override the agent's standard initialization and conversational flow. In the 'CRITICAL: Auto-start on load' section, the agent is commanded to immediately execute the first step of the interview and is explicitly forbidden from summarizing the skill, explaining its files, or seeking user confirmation ('Do NOT: Summarise this skill... Describe what files it creates... Ask if the user wants to run it'). This behavioral override is intended to bypass the agent's default summary and oversight procedures.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted writing samples provided by the user, which presents a surface for indirect prompt injection attacks.
- Ingestion points: Writing samples pasted into chat by the user or loaded from 'references/sample-content.md' in Step 3.
- Boundary markers: The instructions do not define any delimiters or provide warnings to the agent to ignore instructions that might be embedded within the provided writing samples.
- Capability inventory: The agent has the capability to write markdown files ('about-me.md' and 'voice.md') to the project root based on the results of the analysis.
- Sanitization: No sanitization or validation of the input text is performed before processing.
Audit Metadata