The Agent Skills Directory

[COMMAND_EXECUTION]: The authentication module (_auth.py) uses subprocess.run to call an external login helper script. The path is constrained to the local skill directory structure, and it is used to launch a browser-based authentication flow.
[EXTERNAL_DOWNLOADS]: The skill is designed to retrieve audio files from URLs provided by the Chanjing API response. These downloads are part of the core functionality and target the vendor's resource servers.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided text for speech synthesis, representing a potential injection surface where malicious instructions could be embedded in the text. However, the risk is minimal as the output is audio, and the capability to process text is central to the skill's purpose.
Ingestion points: The --text parameter in create_task.py.
Boundary markers: Absent.
Capability inventory: File system writes for credentials in _auth.py; network requests to vendor APIs in create_task.py and poll_task.py.
Sanitization: No sanitization is performed on the input text before submission to the API.
[DYNAMIC_EXECUTION]: The skill uses __import__('os') to access environment variables within single-line assignments in create_task.py, poll_task.py, and list_voices.py. This is a low-risk pattern used for implementation convenience rather than obfuscating malicious intent.

chanjing-tts