The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ffmpeg and sips to process images and verify their properties. These tools are executed via scripts/chroma_cutout.sh and direct instructions in SKILL.md. The bash script correctly employs double-quoting for input and output variables to prevent basic argument injection.\n- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes external image assets and metadata to modify project files such as assets.json and spriteAssets.ts.\n
Ingestion points: Sprite source files, draft paths, and asset metadata provided as inputs to the workflow and helper scripts.\n
Boundary markers: No specific delimiters or safety instructions are defined for the manifest updates or file writes.\n
Capability inventory: File system writes to .webp, .json, .ts, and .md files; execution of local image processing tools and npm build scripts.\n
Sanitization: Shell variables are quoted in the provided script, but no explicit sanitization for manifest data contents is described beyond the inherent constraints of the tools used.

sprite-asset-promotion