The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a large attack surface for indirect prompt injection by accepting free-form text through the --prompt and --lyrics arguments. These inputs are processed by the underlying AI model and could be used to embed malicious instructions.
Ingestion points: User-provided text for music prompts and song lyrics via command-line flags.
Boundary markers: The skill documentation does not provide delimiters or instructions to the agent on how to isolate user content from system instructions.
Capability inventory: The skill utilizes shell execution via uv run to call tools/music_gen.py and performs network operations to interact with the ACE-Step API.
Sanitization: There is no evidence of input validation, escaping, or filtering for the strings provided to the generation tool.
[COMMAND_EXECUTION]: The skill is centered around executing shell commands to interact with its core tool and manage deployments to cloud providers like Modal and RunPod. While this is the intended functionality, it relies on the integrity of the local tools/music_gen.py and the docker/modal-music-gen/app.py script.

acestep