Skill UI

A systematic workflow designed to significantly improve the reliability and performance of existing AI agents. It covers detailed performance analysis, including gathering metrics (e.g., task success rate, latency), classifying failure modes, and advanced prompt engineering techniques (like Chain-of-Thought and Constitutional AI). The process culminates in structured A/B testing and controlled deployment, ensuring measurable, robust improvements.

Provides comprehensive guidance on building robust evaluation frameworks for complex agent systems. It addresses the inherent challenges of non-determinism and multiple valid paths, enabling systematic performance testing, validating context engineering choices, and measuring continuous improvements across multiple quality dimensions (e.g., factual accuracy, tool efficiency).

llama.cpp is a pure C/C++ library designed for efficient LLM inference, optimized to run on resource-constrained environments. It excels in scenarios where high-end NVIDIA GPUs (CUDA) are unavailable, making it ideal for Apple Silicon Macs, AMD/Intel GPUs, or edge deployment devices like Raspberry Pi. It supports GGUF quantization formats, enabling significant memory reduction and speed improvements for local, cross-platform AI applications.

This advanced workflow provides comprehensive pronunciation training by analyzing speech down to the phoneme level. It assesses users' speaking skills, identifies specific weak phonemes (e.g., 'th' vs. 'd'), and runs adaptive drill loops. The system generates detailed weakness reports and recommends targeted practice phrases for continuous accent reduction and fluency improvement.

This skill automates the scholarly process of refining a generated academic paper. It executes an iterative loop (Review → Fix → Recompile) using advanced LLMs (like GPT-5.5) over multiple rounds. It is designed not for primary research, but specifically for polishing writing quality, correcting theoretical inconsistencies, improving presentation, and enhancing narrative flow after the initial draft is complete. It helps elevate a paper from a draft to a polished submission.

This skill automates the research improvement process by creating a closed loop: Review -> Implement Fixes -> Re-review. It leverages any OpenAI-compatible LLM API to iteratively refine research work until a positive assessment is reached or a maximum round limit is hit. Ideal for academic or technical project refinement.

Provides patterns and techniques for building robust evaluation loops for AI agents. This skill guides the implementation of self-critique, evaluator-optimizer pipelines, and test-driven refinement workflows. Use it to significantly enhance the accuracy and quality of complex outputs like code, reports, and analyses by implementing structured, iterative improvement cycles (Generate → Evaluate → Critique → Refine).

Evolver is a comprehensive self-evolution engine designed for autonomous AI agents. It analyzes complex runtime history, identifies performance bottlenecks, and autonomously generates and applies improvements. Communication is architecturally confined to a local Proxy Mailbox, ensuring secure and reliable interaction with the EvoMap Hub for asset management and task acquisition. Ideal for developing highly adaptive, continuously improving intelligent systems.

Simulates a comprehensive, multi-round patent examiner review process. It analyzes patent claims and specifications against international standards (USPTO, CNIPA, EPO) to assess novelty, non-obviousness, claim clarity, and enablement. The tool provides detailed office actions, critical feedback, and iterative improvement rounds to maximize patentability.

This skill provides comprehensive analysis and improvement suggestions for plugin prompts, skill definitions, and command instructions. It evaluates content across five dimensions—clarity, safety, effectiveness, completeness, and conciseness—providing a detailed score card and concrete, actionable rewrites to ensure compliance and optimal AI performance.

SkillOpt-Sleep simulates an agent's nightly 'sleep' cycle to continuously improve performance without requiring model retraining. It harvests historical sessions, mines recurring tasks, and replays them offline. Learning and insights are then consolidated into the agent's memory and skills (CLAUDE.md/SKILL.md), proposing bounded, validated edits that the user must explicitly adopt, ensuring measurable, sustained improvement over time.