The Agent Skills Directory

[SAFE]: The skill defines a set of best practices and development standards for building Reinforcement Learning simulations. It includes detailed architectural patterns (Scene pattern, Router pattern) and coding rules aimed at robustness and testability.
[SAFE]: It explicitly forbids the use of dangerous functions such as eval(), exec(), and getattr() on untrusted keys (Rule 9 in references/code-rules.md), which aligns with secure coding best practices.
[SAFE]: The identified dependencies, including gymnasium, numpy, raylib, python-fire, ruff, and pytest, are standard and reputable libraries within the Python and AI development communities.
[SAFE]: Configuration files for task runners (justfile) and pre-commit hooks (pre-commit-config.yaml) utilize well-known tools and do not contain any suspicious commands, network exfiltration attempts, or unauthorized file access patterns.
[SAFE]: The instruction set focuses on technical implementation details and project structure without any attempts at prompt injection or bypassing safety constraints.

developer