The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests Godot scene files (.tscn) and project configuration files (project.godot), which contain user-controlled identifiers such as node names and property values. These identifiers are extracted and interpolated into audit reports and logs used by the agent for decision-making.
Ingestion points: light_audit.gd and gi_bake.gd load and process external scene files to identify rendering components.
Boundary markers: The skill does not implement specific delimiters or instructions to the LLM to ignore potentially malicious instructions embedded in node names or property strings.
Capability inventory: The skill possesses significant capabilities, including local file writes (saving scenes), image capture, and the execution of Godot scripts via the gd_run toolkit.
Sanitization: No filtering or sanitization is performed on scene-extracted strings before they are presented to the agent's context.
[COMMAND_EXECUTION]: The skill automates complex editor workflows by executing Godot scripts and manipulating the Godot EditorInterface. This includes programmatic interaction with the editor UI to trigger lightmap baking and capturing high-resolution renders for analysis.
[DYNAMIC_EXECUTION]: The post_effect.gd script dynamically loads and compiles GLSL compute shaders at runtime using the Godot RenderingDevice API. This is a legitimate architectural requirement for implementing the skill's custom post-processing effects.

scenario-godot-rendering-lighting