Skill UI

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.

This guide provides a comprehensive methodology for load testing and capacity planning within Snowflake data warehouses. It covers optimizing query performance by adjusting warehouse sizes (Scale UP), simulating high concurrency using multi-threading (Scale OUT), and configuring advanced scaling policies (Standard, Economy, Maximized). Users can benchmark critical queries using SQL and simulate real-world user traffic using Python, ensuring robust and cost-efficient production environments.

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.

A comprehensive reference guide providing quantitative benchmarks and formulas for system architects. It covers user-to-RPS conversion, storage estimation, bandwidth calculation, known scale limits of major technologies (databases, queues, caches), and capacity planning for various user loads (1K DAU to 10M DAU). Essential for designing scalable, high-availability applications.

This comprehensive skill assesses pre-launch readiness across all organizational functions, including Product, Marketing, Support, Sales, and Data. It systematically checks feature completeness, performance benchmarks, operational readiness, and communication assets. The output provides a clear Go/Conditional Go/No-Go recommendation, a prioritized blockers list with owners, a detailed risk register, and a rationale, ensuring a robust and controlled product release.

Transforms complex raw product metrics into a clear, actionable health narrative. This skill analyzes key performance indicators (KPIs) across the entire user lifecycle—including acquisition, activation, engagement, and retention—against set OKRs and benchmarks. The output is a comprehensive report providing overall health status, key observations, hypothesized root causes for performance dips, and prioritized, actionable next steps with assigned owners.

This tool evaluates how thoroughly a feature implementation meets product requirements (PRDs) or specifications. It scores acceptance criteria case-by-case, independently grading both the code implementation and the associated test coverage (unit and E2E). Ideal for benchmarking multiple development efforts, auditing adherence to specifications, or determining if a feature is 100% complete and compliant with the ground truth PRD.

kube-bench is an open-source tool that validates the compliance and security posture of Kubernetes clusters. It rigorously checks control-plane components, worker nodes, and cluster policies against the comprehensive CIS Kubernetes Benchmark. It reports findings (PASS/FAIL/WARN) and includes specific remediation guidance, making it essential for compliance auditing and hardening efforts.

This professional tool helps developers and researchers identify the optimal AI model for a specific task, such as coding, reasoning, or image recognition. It queries official HuggingFace leaderboards, filters models based on hardware constraints (e.g., available VRAM or RAM), and provides a detailed comparison table including benchmark scores, parameter size, and licensing information.

A specialized presentation deck template adopting a cyberpunk, CRT terminal aesthetic. It is ideal for conducting deep technical reviews, such as benchmarking CLI agents, demonstrating code diffs, or presenting complex development tool analyses. Features a dark background, grid lines, command prompt indicators, and structured tagging for maximum professional impact.

This skill executes the DeepSWE benchmark, a comprehensive coding-agent evaluation suite. It uses OpenRouter and mini-swe-agent to score various Large Language Models (LLMs) on 113 tasks, providing reproducible results for model comparison, development assessment, and research benchmarking. Use this when a standardized, deep evaluation of LLM coding capabilities is required.