The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestrator utilizes the Bash tool to run workspace-resident scripts, including helpers/render.py for audio fades, helpers/loudness.py for mastering verification, and hooks/validate-export-name.js for naming convention enforcement. It also facilitates project scaffolding via templates.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external, potentially untrusted content.\n
Ingestion points: Data is ingested from ASR transcripts in workspace/video-projects/<slug>/transcripts/ and project state files (project.md).\n
Boundary markers: The instructions lack explicit guidance for the agent to treat ingested transcript content as untrusted data or use specific delimiters to prevent command execution or behavioral overrides.\n
Capability inventory: The skill has access to sensitive tools including Bash, Write, WebFetch, and the ability to dispatch to other skills via Task and Skill tools.\n
Sanitization: There is no documented process for sanitizing or filtering the content of transcripts before they are utilized by sub-skills in the content production pipeline.

music-studio