voice-builder

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill explicitly orders the agent to auto-start the interview and to withhold any preamble or disclosure to the user (suppressing transparency) and contains a contradictory line forbidding creation of voice.md, which is deceptive relative to the skill's stated outputs and purpose.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Skill reads user-provided “samples” content (pasted writing) in Step 3/4 and uses it for analysis at runtime, so outsider-authored free text can be ingested.