The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for building and interacting with reinforcement learning environments using gymnasium. This creates a processing surface for external data (observations and rewards).\n
Ingestion points: Data is ingested through env.step() and env.reset() calls within training and evaluation scripts.\n
Boundary markers: No explicit boundary markers or instructions to ignore potential commands embedded in environment observations are provided.\n
Capability inventory: The skill includes file system write capabilities via model.save(), vec_env.save(), and video recording in VecVideoRecorder.\n
Sanitization: No specific sanitization or validation of observation data is present in the provided code snippets.\n- [DYNAMIC_EXECUTION]: The skill utilizes model.save() and model.load(). In the context of Stable Baselines3, these functions use Python's pickle module for serializing and deserializing models and environment statistics. Loading model files from untrusted sources could lead to arbitrary code execution due to the nature of pickle deserialization.

stable-baselines3