speech-to-text

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to install standard library dependencies from official registries (PyPI and NPM) for interacting with the ElevenLabs API.
  • Python packages: elevenlabs, python-dotenv, pydub.
  • Node.js packages: @elevenlabs/elevenlabs-js, @elevenlabs/client, @elevenlabs/react, dotenv.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection as it processes and transcribes untrusted audio and video content from user-provided files and external URLs.
  • Ingestion points: SKILL.md and references/realtime-server-side.md contain examples of transcribing local files (audio.mp3) and remote streams (https://npr-ice.streamguys1.com/live.mp3).
  • Boundary markers: The skill examples do not explicitly implement delimiters or instruction-isolation warnings for the transcribed output.
  • Capability inventory: The skill converts audio to text; the security impact depends on the agent's overall tool access when processing the resulting transcription.
  • Sanitization: No transcription sanitization or filtering logic is present in the provided snippets.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 08:42 AM