The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages sensitive API credentials, including app_id, secret_key, and access_token, by reading from and writing to a local configuration file at ~/.chanjing/credentials.json. This is standard practice for the vendor's credential management.
[COMMAND_EXECUTION]: The _auth.py script invokes subprocess.run to execute a local script from a sibling directory (chanjing-credentials-guard/scripts/open_login_page). This is used as an automated fallback to open a login page if credentials are missing.
[EXTERNAL_DOWNLOADS]: The skill makes several network requests using urllib.request to https://open-api.chanjing.cc. These requests are used to obtain access tokens, upload reference audio for voice cloning, and poll for the status of speech synthesis tasks.
[PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection by processing user-supplied text and URLs for TTS processing.
Ingestion points: User-provided text and url inputs in create_task and create_voice scripts.
Boundary markers: None identified; input is interpolated directly into JSON request bodies.
Capability inventory: File system access (read/write credentials), subprocess execution (_auth.py), and network operations across all primary scripts.
Sanitization: The create_task script enforces a maximum length of 4000 characters for the input text.

chanjing-tts-voice-clone