The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using subprocess.run to call ffmpeg and ffprobe for audio concatenation, format conversion, and duration probing in several backend scripts, such as scripts/backends/azure.py, scripts/backends/baidu.py, and scripts/backends/edge.py.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to communicate with multiple third-party TTS providers, including Microsoft Azure, OpenAI, Google Cloud, and ElevenLabs. It also supports downloading reference audio from remote URLs for voice cloning in scripts/clone.py.
[DATA_EXFILTRATION]: The skill transmits user-provided text or the contents of files to external TTS service providers for audio synthesis as part of its core functionality.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to the processing of untrusted input data:
Ingestion points: User-supplied text enters through the text argument or the --input file flag in scripts/tts.py.
Boundary markers: The skill does not implement boundary markers or instructions to the synthesis backends to disregard control characters or malicious instructions within the input text.
Capability inventory: The skill can execute local shell commands (ffmpeg) and perform network requests to external APIs.
Sanitization: Input text is processed and sent to external services without sanitization or filtering for potentially malicious content.

ttscn