Advisory MCP gate that probes coding-agent tool-calls via a Llama-3.1-8B SAE inspector. Observation only, fail-closed escalate, sandbox required.
-
Updated
May 25, 2026 - Python
Advisory MCP gate that probes coding-agent tool-calls via a Llama-3.1-8B SAE inspector. Observation only, fail-closed escalate, sandbox required.
Lightweight exploratory toolkit inspired by Goodfire-style neural-geometry experiments — find circular representations, Fourier features, and modular-arithmetic geometry in LLM activations. Not affiliated with Goodfire.
Add a description, image, and links to the goodfire topic page so that developers can more easily learn about it.
To associate your repository with the goodfire topic, visit your repo's landing page and select "manage topics."