The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill is designed to write music as code in score.mjs and subsequently execute it to render audio files. While this is the core functionality, generating and running scripts at runtime carries inherent risks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data, including narration text and existing project files like cue.md, analysis.json, and music.meta.json.
Ingestion points: The playbooks revise-cue.md and create-cue.md instruct the agent to read project metadata, previous scores, and user-provided narration or timeline data.
Boundary markers: No specific delimiters or warnings to ignore embedded instructions are defined for the ingested content in the prompts.
Capability inventory: The skill has the ability to write files, execute generated JavaScript code, and run shell commands via sibling tools (e.g., op55-render, op55-verify).
Sanitization: There is no explicit instruction to sanitize, escape, or filter the external content before processing it.
[COMMAND_EXECUTION]: The workflow involves executing various shell commands for audio processing, rendering, and verification. The specific commands and script paths are retrieved from a sibling skill (resolve-op55-tools).

op55-mode