The Agent Skills Directory

[SAFE]: The skill references the belt CLI tool from inference.sh, which is a well-known service for AI-driven audio and media inference tasks.
[SAFE]: External configuration and installation files are sourced from official repositories within the inference-sh GitHub organization.
[SAFE]: The skill follows secure credential management practices by instructing users to authenticate using the belt login command instead of requiring hardcoded secrets or environment variables.
[PROMPT_INJECTION]: The skill processes audio content from external URLs provided as input. This constitutes an indirect prompt injection surface, as the agent might ingest untrusted data that could potentially contain embedded instructions intended to influence its behavior.
Ingestion points: SKILL.md (via --input parameters in CLI command examples).
Boundary markers: No specific boundary markers or delimiters are used to separate user-provided audio data from the agent's control context.
Capability inventory: The skill utilizes the belt CLI tool within a Bash environment to perform audio processing.
Sanitization: There is no evidence of validation or sanitization for the input URLs or the content they retrieve.

elevenlabs-voice-isolator