The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bridge script (gd_run) to execute Godot engine commands for headless layout auditing and scene captures. This allows the agent to verify UI behavior across various target resolutions without requiring a graphical display.
[DYNAMIC_EXECUTION]: The toolkit dynamically loads and instantiates GDScript files using Godot's internal ResourceLoader and PackedScene systems, particularly within ui_selftest.gd and ui_layout.gd, to verify project resources and UI hierarchies at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data, creating an attack surface for indirect prompt injection.
Ingestion points: Localization data is ingested via CSV files in scripts/gd_ui.py, and Godot scenes are instantiated for reporting in scripts/agentkit/ui/ui_layout.gd.
Boundary markers: There are no specific instructions or delimiters used to warn the agent about ignoring embedded instructions within the processed data.
Capability inventory: The skill maintains high-privilege capabilities including filesystem writes to generate project scripts (build_g7.gd) and the ability to modify project settings (ui_settings.gd).
Sanitization: The check_csv function in the Python helper script performs structural validation (placeholder parity, BBCode balance) but does not specifically filter for natural language instruction injection.

godot-ui