speech-to-text

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious patterns or security vulnerabilities were identified in the skill's instructions or referenced files.
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing official ElevenLabs packages (elevenlabs for Python, @elevenlabs/elevenlabs-js for Node.js) and standard utility libraries (pydub, dotenv) from public package registries. These are legitimate dependencies required for the stated functionality.
  • [CREDENTIALS_UNSAFE]: The documentation correctly instructs users to manage the required ELEVENLABS_API_KEY using environment variables. Code examples use placeholders or environment variable calls rather than hardcoding sensitive credentials.
  • [COMMAND_EXECUTION]: The documentation includes standard CLI commands for package management and API testing via curl. These are standard developer tools and do not represent unauthorized or dangerous command execution.
  • [PROMPT_INJECTION]: The skill processes external audio and video content for transcription. While this is the primary function, it presents an indirect prompt injection surface where malicious instructions embedded in the audio could potentially influence the agent's behavior after the text is processed.
  • Ingestion points: Transcription of local files (SKILL.md) and remote URLs (references/realtime-server-side.md).
  • Boundary markers: None specified in the provided code snippets.
  • Capability inventory: Converts audio/video streams to text via the ElevenLabs Scribe API.
  • Sanitization: No explicit sanitization of transcribed text is shown in the examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 02:48 AM
Security Audit — agent-trust-hub — speech-to-text