The Agent Skills Directory

[Data Processing Surface]: The skill ingests and processes image assets, such as seed frames, to create animation sequences. Processing external files is a point where the agent interacts with data that could potentially contain indirect instructions.
Ingestion Points: The "Core Workflow" in SKILL.md initiates processing using an "approved in-game seed frame."
Boundary Markers: The instructions do not currently include specific markers or guidelines to disregard instructions that might be embedded in asset metadata.
Capability Inventory: The skill performs file writing and image manipulation via local script execution (build_sprite_edit_canvas.py, normalize_sprite_strip.py, render_sprite_preview_sheet.py).
Sanitization: There are no explicit instructions for validating or sanitizing the input assets before processing.
[Local Pipeline Execution]: The skill defines a series of shell commands for local Python scripts to automate the asset generation process. These commands manage the flow of data between generation and normalization steps.
Evidence: SKILL.md includes recipes for running scripts like normalize_sprite_strip.py and render_sprite_preview_sheet.py to handle project assets.

sprite-pipeline