The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion and processing of untrusted external audio content, creating an attack surface for indirect prompt injection if the resulting transcripts contain malicious instructions.\n
Ingestion points: The audioUrl parameter used in the transcribeWithBackoff, queuedTranscribe, and batchTranscribe functions in SKILL.md.\n
Boundary markers: Absent; the code snippets and instructions do not include or recommend the use of delimiters or 'ignore' warnings for the transcribed output.\n
Capability inventory: The implementation utilizes the assemblyai SDK to perform network requests to an external transcription service.\n
Sanitization: Absent; the instructions do not provide methods for validating input URLs or filtering the transcription text for potentially harmful content.

assemblyai-rate-limits