Skill UI

This harness provides comprehensive, reproducible evaluation pipelines for complex LLM chains and agents (LangChain/LangGraph 1.0). It integrates golden dataset management, LangSmith evaluation, RAGAS metrics, deepeval LLM-as-judge, and structured agent trajectory analysis. Use it when establishing quality benchmarks for new chains, diagnosing performance regressions after model switches, or implementing CI/CD gates to prevent quality drops.

This comprehensive benchmark suite evaluates the performance of Vision-Language Models (VLM) on complex video anomaly detection tasks using the SmartHome-Bench dataset. It covers a wide range of real-world smart home scenarios, including security breaches, elderly fall detection, and unusual pet behavior. Unlike single-frame analysis, this benchmark requires deep multi-frame video understanding, making it a robust tool for testing the advanced AI capabilities of video monitoring systems.

Combines a curated Japanese NLP resource dataset with live web research to produce an up-to-date trend report on models, tooling, benchmarks, embeddings, speech, and related subfields.

This skill implements Autonomous Optimization (AO) using Hypothesis Tree Refinement (HTR). It iteratively improves concrete artifacts (like code, models, or data pipelines) by running structured, long-term experiments against a measurable objective. It maintains a persistent hypothesis tree, coordinating subagents to test hypotheses in isolation while using a held-out test evaluator to prevent overfitting and ensure reliable, cumulative research progress. Ideal for complex benchmark tuning, agent refinement, or optimizing entire workflows.