The Agent Skills Directory

[SAFE]: The skill defines a structured workflow for generating instrumental music and vocal songs using an internal tool. It contains specific instructions to avoid duplicating lyrics, imitating named artists, or using copyrighted recordings, which aligns with security and ethical best practices.
[INDIRECT_PROMPT_INJECTION]: The skill contains logic for ingesting untrusted user data via the lyrics and prompt fields in SKILL.md and references/song.md.
Ingestion points: User-supplied lyrics and style descriptions are processed by the submit_music tool.
Boundary markers: The instructions recommend using structure markers like [Verse] and [Chorus] to delimit content.
Capability inventory: The skill uses submit_music for audio generation and track_progress for status monitoring; no direct file system writes or arbitrary network operations are performed by the skill logic itself.
Sanitization: The skill instructs the agent to preserve user lyrics as-is and seek approval for drafted lyrics, relying on the underlying generation tool's internal safety filters for the generated audio output.

music