The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of audio URLs and user-provided lyrics which are interpolated into prompts for music generation models. Maliciously crafted audio files or steganographic content could potentially influence the model's output or the agent's behavior during processing.
Ingestion points: The audio, prompt, tags, and lyrics parameters in the input JSON for Routes 1 through 4.
Boundary markers: The skill relies on passing inputs via a JSON string through the CLI's --input flag, which protects against shell-level injection but does not prevent instruction following from the content of the data.
Capability inventory: The skill possesses file-writing capabilities (downloading audio to ./out), command execution (runcomfy CLI), and network access to the RunComfy API.
Sanitization: The skill includes a 'Security & Privacy' section that explicitly identifies audio URLs as untrusted and advises the agent to only ingest content explicitly provided by the user.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the runcomfy CLI using standard package managers (npm, npx). These are official vendor resources from 'runcomfy-com'. The documentation proactively warns against insecure installation methods like piping remote scripts directly into a shell.
[COMMAND_EXECUTION]: The skill utilizes the runcomfy CLI tool to interact with AI music models. The allowed-tools configuration correctly limits the agent's scope to the runcomfy command, adhering to the principle of least privilege.

ai-music