elevenlabs-core-workflow-a

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data to be converted into speech, which represents a potential surface for indirect prompt injection if the processed content is later re-ingested by an agent.
  • Ingestion points: The text parameter in the generateSpeech function (SKILL.md) and the textChunks array in streamTTSWebSocket (references/implementation.md) serve as the primary entry points for external data.
  • Boundary markers: The implementation does not demonstrate the use of boundary markers or delimiters to isolate the text being synthesized from other agent instructions.
  • Capability inventory: The skill utilizes network capabilities via the @elevenlabs/elevenlabs-js library and WebSockets (ws) to communicate with ElevenLabs APIs, and it performs file system write operations to save audio output.
  • Sanitization: There is no evidence of text sanitization or filtering to prevent the inclusion of malicious instructions within the synthesized text.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 02:20 PM
Security Audit — agent-trust-hub — elevenlabs-core-workflow-a