elevenlabs-stt
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill facilitates audio transcription and forced alignment using the ElevenLabs Scribe models via the
beltCLI tool. All external references and commands are directed toward the author's official infrastructure (inference-sh). - [PROMPT_INJECTION]: The skill ingests untrusted data in the form of audio URLs and text strings for processing. This represents a standard indirect prompt injection surface common to speech-to-text applications.
- Ingestion points: Audio URLs and text inputs are passed to the
belt app runcommand withinSKILL.md. - Boundary markers: None present; the skill treats the input as standard JSON payloads.
- Capability inventory: The skill is limited to the
Bash(belt *)tool, restricting execution to the vendor's CLI. - Sanitization: No explicit sanitization or filtering of the audio transcription output is performed before it returns to the agent context.
Audit Metadata