The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a search routine to locate and execute Node.js scripts (image-util.mjs and pack-spritesheet.mjs) from other installed skills (pixel-snapper and animated-spritesheets) using computed filesystem paths.
[COMMAND_EXECUTION]: The instructions explicitly guide the agent to structure shell commands to evade the platform's permission prompt mechanisms, stating that this allows the tool to run without user-authorized prompts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (character descriptions) and interpolates it directly into model prompts without using boundary markers or input validation, creating a surface for indirect prompt injection.
Ingestion points: Character descriptions in Recipe 1 and throughout the pipeline.
Boundary markers: Absent.
Capability inventory: Uses the vg tool for network-based asset generation and file writes via the --download flag.
Sanitization: None described in the prompt templates.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install a global NPM package (vibedgames) and uses the vg CLI to download generated assets from remote CDNs to the local filesystem.

pixel-art