The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Unreal Engine projects, including material graph structures, texture metadata, and asset names, which represents a potential indirect prompt injection surface.
Ingestion points: audit_textures, export_graph, and census_project in scripts/ue_materials.py read asset names and properties from the project registry.
Capability inventory: The skill has the ability to create and modify assets, save changes to the filesystem, and execute Unreal Engine console commands.
Boundary markers: Findings and reports generated from asset data do not use specific delimiters to isolate potentially untrusted asset names.
Sanitization: Input data is validated for technical correctness according to Unreal Engine standards but is not specifically sanitized for instruction-like patterns.
[DYNAMIC_EXECUTION]: The skill uses dynamic code generation and execution to programmatically build material graphs and interact with the Unreal Engine editor.
Node generation: scripts/ue_materials.py includes functions like build_kit_surface that use the MaterialEditingLibrary to instantiate and link material nodes at runtime.
Remote execution: Procedure P9 in references/procedures.md utilizes the ue_remote toolkit to execute Python commands within a live Unreal Engine editor instance for performance profiling.
Shader injection: Procedure P13 in references/procedures.md generates HLSL shader code and injects it into MaterialExpressionCustom nodes.
[COMMAND_EXECUTION]: The skill invokes Unreal Engine console commands for editor configuration and performance analysis.
Execution method: The console() function in scripts/ue_materials.py calls unreal.SystemLibrary.execute_console_command to set CVars and trigger engine features like HighResShot and CsvProfile.

unreal-materials