Skill UI

A sophisticated, two-stage system designed to capture and internalize user corrections and learnings during ongoing conversations. It automatically detects patterns (e.g., "use X instead of Y") and queues them. Users can manually review and apply these learnings to a persistent knowledge base (CLAUDE.md), ensuring the AI continuously improves its knowledge and adherence to specific user context for future sessions.

Guides researchers through nnsight’s tracing and activation APIs so you can inspect or modify PyTorch internals locally or via NDIF-powered remote runs on 70B+ foundation models without local GPUs.

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.

This skill implements the BDI (Belief-Desire-Intention) architecture for advanced cognitive agent modeling. It allows agents to reason about context by transforming external knowledge (like RDF triples) into internal mental states (beliefs, desires, intentions) and vice versa. Key features include goal-directed planning, explainable reasoning chains, and T2B2T (Triples-to-Beliefs-to-Triples) flow, making it ideal for multi-agent systems and augmenting LLMs with formal cognitive structures.

This skill acts as an advanced extension layer, implementing a self-evolution protocol, maintaining session state continuity (Compaction), and providing KPI-style performance summaries. It allows users to simulate continuous professional development, track metrics, and master complex role-playing and interaction skills across multiple sessions, ensuring performance baselines are continuously adjusted and internalized.

A comprehensive security auditing framework designed to systematically detect sensitive information disclosure from Large Language Models (LLMs). It uses multi-phase, escalating dialogue probes to test for leaks such as system prompts, API keys, PII, and internal configurations. Essential for red teaming and rigorous model security assessment.

A specialist tool designed to pressure-test AI Management Systems (AIMS) against the ISO 42001 standard. It executes six critical compliance checks, covering scope definition, risk mapping, policy adherence, and framework integration. Essential for pre-certification readiness, annual internal audits, and new system onboarding.

A specialized quality gate skill designed to review complex, multi-source outputs generated by the CrossFrame framework. It rigorously checks reasoning fidelity, evidence boundaries, source anchors, and structural continuity. Use this when advanced verification of logical rigor and source integrity is required, ensuring outputs meet strict internal protocols and avoid common structural failures.

This tool provides deep inspection and auditing capabilities for GEPA (genetic-evolution prompt-adaptation) genomes by utilizing the internal `@metaharness/darwin/gepa` library. It is crucial for advanced AI development pipelines, allowing users to validate the structural integrity of genomes, render the full system prompt derived from a genome, or classify specific failure modes found in an execution transcript. Ideal for CI/CD gating and debugging complex LLM applications.