The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes a build(code=...) tool which executes Python code strings at runtime to generate procedural landscapes. While the provided examples are benign noise functions, this represents a capability for dynamic code execution.
[INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by documenting how an agent should generate and execute code or math expressions (via build or worldedit_deform) to fulfill terrain requests.
Ingestion points: The agent is expected to translate natural language terrain descriptions into code strings for the build() tool.
Boundary markers: No specific delimiters or safety instructions are provided to prevent user-supplied descriptions from injecting malicious logic into the generated Python code.
Capability inventory: The build() tool allows Python execution, and WorldEdit tools allow executing arbitrary math expressions and Minecraft server commands (e.g., /setblock).
Sanitization: The skill lacks guidance on sanitizing or validating descriptions before they are interpolated into executable code or command templates.

generating-terrain