The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the runcomfy CLI tool via the Bash tool to interact with AI music models. The execution is restricted to the runcomfy command prefix as defined in the allowed-tools frontmatter.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface when processing third-party audio files for editing tasks. It provides a mandatory evidence chain for this risk:
Ingestion points: The audio URL parameter used in the audio-inpaint and audio-outpaint routes in SKILL.md allows external data into the agent's workflow.
Boundary markers: The instructions mandate passing all prompt, lyric, and URL data within a structured JSON object via the CLI's --input flag to separate data from shell instructions.
Capability inventory: The skill utilizes the runcomfy CLI (invoked via Bash) to perform file downloads and API interactions.
Sanitization: The documentation notes that the CLI transmits the JSON body directly to the Model API over HTTPS without performing shell expansion on the content.

ai-music