elevenlabs
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data for text-to-speech, sound effects, and music generation, creating a surface for indirect prompt injection attacks.\n
- Ingestion points: The
textandpromptparameters in audio generation functions withinSKILL.mdandreference.md(e.g.,client.text_to_speech.convert(text=...)), as well as scene metadata ingested frommanifest.jsonin the Remotion integration section.\n - Boundary markers: Absent. No delimiters or specific instructions (e.g., 'ignore embedded commands') are used when interpolating text into API calls.\n
- Capability inventory: The skill performs network requests to ElevenLabs APIs, reads local files for voice cloning, and writes generated audio data to the local filesystem using
save()and file write operations.\n - Sanitization: Absent. There is no evidence of input validation or escaping for the text content before it is processed by the audio synthesis models.
Audit Metadata