voice
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to manipulation through untrusted input data that is processed for text-to-speech (TTS) generation. Malicious instructions or functional tags embedded in source text could influence the audio output's tone or delivery.
- Ingestion points: The
textparameter in thesubmit_voicetool call receives content that may be derived from external scripts, storyboard beat descriptions, or visual analysis results (SKILL.md, references/video-sync.md). - Boundary markers: No explicit delimiters or instructions are provided to the agent to sanitize or distinguish intended narration text from potentially malicious functional tags (e.g.,
[happy],[whispers],[pause]) when processing third-party content. - Capability inventory: The skill uses the
submit_voiceandsubmit_soundtools to perform network-based media synthesis via external providers (Doubao, ElevenLabs, Fish Audio). - Sanitization: The instructions do not include steps to filter or validate control tags within the input text before passing them to the synthesis tools.
Audit Metadata