The Agent Skills Directory

[DATA_EXFILTRATION]: Accesses a sensitive configuration file at ~/.env via the load_env function in scripts/tts.py. This file often contains credentials for various services, and reading it from the home directory constitutes a data exposure risk.
[COMMAND_EXECUTION]: Executes the ffmpeg CLI tool using subprocess.run to convert WAV files to OGG format. The implementation correctly uses an argument list which mitigates shell injection risks.
[EXTERNAL_DOWNLOADS]: Fetches AI models from Hugging Face, a well-known service, when TTSModel.load_model() is called. These downloads are necessary for the primary function of the skill.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection where untrusted text input could influence agent behavior if the output is processed by other skills.
Ingestion points: The text argument in scripts/tts.py accepts arbitrary strings for speech generation.
Boundary markers: Absent; no delimiters are used to wrap or isolate the processed text input.
Capability inventory: File system writing, subprocess execution (ffmpeg), and network access for model downloads.
Sanitization: No sanitization or validation is applied to the input text before processing.

ez-tts