The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill utilizes RUNPOD_API_KEY and RUNPOD_ACESTEP_ENDPOINT_ID for authentication with the music generation backend. It correctly follows security best practices by instructing users to store these credentials in a .env file, preventing accidental exposure of secrets in the instruction text or command history.
[INDIRECT_PROMPT_INJECTION]: The skill includes a standard attack surface for indirect prompt injection, which is inherent to media generation tools that process external content.
Ingestion points: The skill processes untrusted user data via text prompts (--prompt), song lyrics (--lyrics), and reference audio files (--reference, --input).
Boundary markers: The provided documentation does not specify the use of delimiters or specific boundary markers to isolate user-supplied inputs from the tool's control parameters.
Capability inventory: The skill invokes tools/music_gen.py, which is capable of making network requests to the RunPod API and performing local file system write operations (--output).
Sanitization: The documentation does not describe internal sanitization or validation mechanisms for the text and audio inputs ingested during the music generation process.
[SAFE]: The skill relies on local script execution (tools/music_gen.py) and does not perform any suspicious external downloads, remote code execution from untrusted sources, or persistence-related activities.

acestep