The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/generate_tts.py script automates system utilities such as ssh, scp, curl, ffmpeg, and ffprobe using subprocess.run. The implementation uses argument lists and avoids shell=True, which is a best practice for preventing command injection.
[CREDENTIALS_UNSAFE]: Sensitive SSH passwords are managed securely by injecting them into the environment of the expect utility rather than writing them to temporary files. The script also includes a redaction step to remove passwords from any captured process output before it is logged.
[DYNAMIC_EXECUTION]: The skill dynamically generates Tcl/Expect scripts to handle interactive authentication. It employs json.dumps for parameter serialization and specific Tcl character escaping to ensure that user-supplied input cannot be interpreted as commands.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user input for speech text and voice descriptions. The risk of these inputs influencing agent behavior is mitigated by multi-layer encapsulation, including JSON serialization and Base64 encoding of the data before it is processed by the remote instance.

qwen3-tts