The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests external audio and video data, transcribes it, and provides the resulting text to the agent.
Ingestion points: The transcribe function in transkription_core.py reads local files provided via user input.
Boundary markers: There are no explicit markers or instructions to the agent to treat the transcribed output as untrusted data.
Capability inventory: The skill allows reading local files and writing to a local SQLite database (store.db).
Sanitization: The transcribed text is returned to the agent without any sanitization or validation of its content.
[EXTERNAL_DOWNLOADS]: The skill metadata identifies openai-whisper and vosk as optional Python dependencies. These are established libraries for speech-to-text processing and are considered safe dependencies for this use case.

transkription