The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation describes the automated download of several AI models during initial command execution. This includes the Kokoro-82M model (~311 MB) for text-to-speech, various Whisper models (up to 3.1 GB) for transcription, and u2net_human_seg weights (~168 MB) for background removal.
[COMMAND_EXECUTION]: The instructions require the agent to execute multiple shell commands. These include using npx hyperframes for media tasks, pip install kokoro-onnx soundfile for Python-based dependencies, and system-level installations via brew or apt-get for the espeak-ng phonemizer.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, which creates a vulnerability surface for indirect prompt injection.
Ingestion points: The tts command accepts arbitrary text strings or text file contents as input, while the transcribe command processes external audio and video files.
Boundary markers: There is no evidence of boundary markers or specific instructions for the agent to ignore or escape potentially malicious instructions embedded within the user-provided text or media files.
Capability inventory: The skill's environment supports command-line execution and file system writes to both output paths and a local cache directory (~/.cache/hyperframes/).
Sanitization: The skill lacks explicit sanitization or validation steps for input data before it is passed to the underlying processing tools.

hyperframes-media