The Agent Skills Directory

[SAFE] (SAFE): No malicious patterns, prompt injections, or obfuscation techniques were detected in the skill files.
[DATA_INGESTION] (SAFE): The training script scripts/train_neural_network.py uses yaml.safe_load() to process configuration files, preventing potential YAML deserialization attacks.
[DYNAMIC_EXECUTION] (LOW): While the script dynamically constructs model architectures from YAML input, it does so using a strict whitelist of supported layer types (e.g., 'Conv2D', 'Dense', 'Flatten'). It avoids dangerous functions like eval() or exec().
[EXTERNAL_DOWNLOADS] (SAFE): The skill references pre-trained model downloads from Hugging Face and PyTorch Hub. These are industry-standard trusted sources for machine learning weights and do not constitute a security risk in this context.

deep-learning