Skill UI

Pyvene is a declarative framework designed for performing causal interventions on PyTorch neural networks. It allows researchers to conduct advanced experiments such as activation patching, causal tracing (ROME-style), and interchange intervention training (IIT). Use this library when you need to test causal hypotheses about model behavior, deeply interpret model components, or ensure reproducibility in advanced AI research.

TransformerLens is the standard library for performing mechanistic interpretability research on large language models. It provides clean interfaces to inspect and manipulate transformer internals—such as attention patterns, residual streams, and MLP outputs—via activation caching and HookPoints. This tool is essential for reverse-engineering model algorithms, performing causal tracing, and studying the internal circuits responsible for model behavior.

A specialized module containing Yann LeCun's comprehensive arguments on the limitations of Large Language Models (LLMs). It critically examines the concept of 'emergent reasoning,' contrasts LLMs' statistical prediction with genuine world models, and addresses deep technical rivalries (e.g., LeCun vs. Hinton). Ideal for advanced discussions on AI causality, common sense, and AGI feasibility.

Query and manage agent knowledge through the controller bridge. Supports semantic routing, hierarchical recall, causal graphs, and context synthesis. Use MCP tools to store, retrieve, search patterns, and build knowledge graphs for AI agents.

A specialized guide for authors aiming to submit to the Conference on Uncertainty in Artificial Intelligence (UAI). It provides expert advice on framing the core contribution—such as probabilistic reasoning, causal inference, and uncertainty modeling—and calibrating the manuscript for highly technical, statistical ML reviewers. Use this skill to refine your submission strategy, identify evidence gaps, and properly position your work relative to other top-tier AI venues like AISTATS and IJCAI.

This skill provides a comprehensive audit of legal AI outputs, documents, and prompts before they are trusted or used in professional contexts. It checks for seven critical 'fault lines,' including grounding, causality, metacognition, and adherence to authority boundaries. Use this for legal memos, contract reviews, compliance reports, or any AI-generated content requiring high reliability.

A specialized module containing Yann LeCun's comprehensive arguments on the limitations of Large Language Models (LLMs). It critically examines the concept of 'emergent reasoning,' contrasts LLMs' statistical prediction with genuine world models, and addresses deep technical rivalries (e.g., LeCun vs. Hinton). Ideal for advanced discussions on AI causality, common sense, and AGI feasibility.

Answer questions about a past agent run from its recorded transcript instead of memory. Find sessions, build a step timeline with recorded vs inferred causality, and verify summaries with typed claim checks.