Skill UI

A comprehensive guide to configuring and maximizing the potential of Claude Code, an advanced agentic coding tool. This resource details best practices, setup templates (CLAUDE.md), advanced prompting techniques (e.g., thinking keywords), and debugging strategies for building reliable, complex software autonomously.

Addresses the critical challenge of managing extremely long conversation histories (millions of tokens) in advanced AI agents, such as coding assistants. This technique moves beyond simple token minimization by implementing structured compression strategies (like Anchored Iterative Summarization) to prevent the loss of critical technical details—such as file paths, decisions, and error messages—ensuring the agent maintains functional context across complex, multi-step tasks. It emphasizes optimizing for 'tokens-per-task' rather than just tokens-per-request.

This tool optimizes the codebase indexing within Cursor. It processes your entire codebase, converting code chunks into vector embeddings stored in a vector database. This enables advanced @Codebase semantic search, allowing users to ask natural language questions (e.g., "Where is authentication handled?") and retrieve the most relevant code snippets, significantly improving AI context awareness across chat and coding tasks.

A comprehensive guide to avoiding common pitfalls when using AI-powered IDEs like Cursor. It covers best practices for AI interaction, secure coding habits (handling secrets, credentials), project configuration (rules, ignore files), and performance optimization. Essential for maintaining high code quality and security in AI-assisted workflows.

Feature Engineering Toolkit automates creating, selecting, and transforming dataset features to improve machine learning models; use it whenever you need new interactions, automatic selection, scaling, or encoding to boost accuracy and interpretability.

A comprehensive toolkit for evaluating the performance of code generation models. It supports benchmarking against industry standards like HumanEval, MBPP, and MultiPL-E across multiple languages. Use this tool to quantitatively compare different LLM coding abilities and measure code synthesis quality using pass@k metrics.

Extends transformer context windows using RoPE, YaRN, ALiBi, and position interpolation so long documents (32k-128k+ tokens) can be processed, pre-trained models lengthened, and efficient positional encodings or extrapolation strategies deployed.

This guide details advanced methods for merging multiple specialized LLMs (e.g., combining math, coding, and chat capabilities) without the need for expensive retraining. Learn various techniques such as Linear Merging, SLERP, Task Arithmetic, TIES-Merging, and DARE. Use mergekit to create high-performing, specialized model variants quickly, significantly improving performance and reducing computational costs.

This skill provides production-ready patterns for integrating with the Perplexity Sonar API in both TypeScript and Python. It covers best practices for creating robust wrappers around the standard OpenAI client, including dedicated handling for citations, search results, and related questions. It also includes essential utilities like implementing exponential backoff for reliable retries. Use this when setting up complex search-augmented generation workflows or standardizing team coding practices around Perplexity usage.

A persistent, file-based planning system for AI coding agents. It maintains task plans, findings, and progress on disk to prevent state loss during context resets, supporting seamless session recovery.

A comprehensive collection of battle-tested, structured prompts designed to elevate interactions with Large Language Models (LLMs). Covers advanced use cases including expert coding reviews, system architecture design, performance analysis, technical writing, and creative brainstorming, ensuring high-quality, actionable AI output.

Speeds up LLM inference by combining speculative decoding with Medusa multiple heads and lookahead techniques so chatbots, code assistants, and other real-time workloads can run with 1.5-3.6× lower latency while preserving target-model quality.