The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted audio data and converts it into text, which serves as a potential vector for indirect prompt injection if the transcribed content contains malicious instructions.
Ingestion points: Transcription text returned from the POST /v1/audio/transcriptions endpoint (SKILL.md).
Boundary markers: Absent. The instructions do not define delimiters or instruct the agent to treat the transcribed output as untrusted content.
Capability inventory: The skill is designed to read local files (node:fs) and transmit data to a remote API via fetch or curl (SKILL.md).
Sanitization: Absent. There are no guidelines for validating or sanitizing the transcribed text before it is consumed by the agent.

9router-stt