The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external project specifications and heightmap files. This data is used to derive world parameters and generate assets via engine commandlets.
Ingestion points: Processes JSON specifications (e.g., valley_spec.json) via json.load and heightmap data (PNG/RAW) via read_png16 in scripts/ue_world.py.
Boundary markers: The skill relies on structured data formats but lacks explicit prompt delimiters or instructions to ignore embedded commands within the processed data.
Capability inventory: The skill has the ability to write files to the local system (write_png16, write_r16, json.dump) and execute Unreal Engine commandlets and Python scripts using the ue_run utility.
Sanitization: It performs structural validation of landscape layouts and height ranges in check_world_spec, but does not sanitize data before it influences script parameters or commandlet arguments.
[COMMAND_EXECUTION]: The skill executes shell-level Unreal Engine commandlets (e.g., WorldPartitionBuilderCommandlet) and Python scripts using the ue_run and ue_run.run_commandlet utilities for building landscapes, PCG content, and HLODs.
[EXTERNAL_DOWNLOADS]: The documentation in SKILL.md instructs the agent to suggest installing sibling skills from the vendor's repository (scenario-labs/skills) using npx if they are missing. These are legitimate vendor resources and align with the skill's modular design.

scenario-unreal-world-building