The Agent Skills Directory

[SAFE]: The skill serves as a reference guide for the TransformerLens library. It provides tutorials on activation caching, patching, and circuit analysis.
[SAFE]: External dependencies such as transformer-lens and torch are standard, well-known packages in the machine learning and interpretability community.
[SAFE]: Instructions for setting environment variables like HF_TOKEN are standard practice for accessing gated models on Hugging Face and do not involve credential exfiltration or hardcoded secrets.
[SAFE]: All external links point to official documentation, public research papers, and established community resources (e.g., GitHub, Google Colab, and the ARENA curriculum).

transformer-lens-interpretability