The Agent Skills Directory

[SAFE]: The skill instructions define high-level logic for procedural generation, emphasizing reproducibility and validation of progression-critical paths.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns detected. References to Godot and Tiled documentation are for informational purposes and point to well-known, trusted domains.
[PROMPT_INJECTION]: No prompt injection or behavior override attempts were found. The skill instructions encourage following best practices for generator stability and error handling.
[DATA_EXFILTRATION]: No network operations or sensitive file access patterns were identified. The skill operates on mathematical seeds and geometric representations.
[OBFUSCATION]: No obfuscated content, encoded strings, or hidden text patterns were detected in the skill or its assets.
[INDIRECT_PROMPT_INJECTION]: While the skill defines how to handle external inputs (seeds, player edits), it lacks executable capabilities like file system writes or network requests that could be exploited via indirect injection. The adversarial test case in evals.json is designed to verify the agent's resistance to bad logic, not to perform an attack.

procedural-generation