The Agent Skills Directory

[SAFE]: The skill uses project-relative paths and avoids machine-specific details, reducing the risk of unintended data exposure. It operates within the scope of the project files provided.
[INDIRECT_PROMPT_INJECTION]: The skill processes external inputs such as generation briefs and prompt files that could contain embedded instructions.
Ingestion points: Workflow steps read WORKFLOW_SELECTION.md, prompt files, and generation briefs.
Boundary markers: The instructions do not define specific delimiters for untrusted content.
Capability inventory: The skill has file read/write capabilities and interacts with generation tools like comfyui and prompt-qc.
Sanitization: The skill mitigates risks through prompt-qc and live-node schema validation during the setup and execution phases.

comfyui-handoff