Skill UI

This guide outlines modern best practices for building clean, maintainable, and scalable user interfaces using Avalonia. It emphasizes leveraging semantic containers (like HeaderedContainer, EdgePanel), defining shared styles, and creating generic components to minimize XAML redundancy and deep nesting. It guides developers to use Behaviors over code-behind, ensuring adherence to clean architectural principles.

Guides developers through configuring Cursor workspaces that span multiple repositories or packages, covering strategy selection, .cursorrules inheritance, selective indexing, and cross-project context handling to keep monorepo-style workflows productive.

Guides developers through mastering Cursor IDE tab completion and AI code suggestions, explaining how to trigger, accept, and configure ghost text workflows for consistent code style and context-aware results.

Guides developers through Exa neural-search anti-patterns, covering query style, search types, content retrieval, date filters, and autoprompt cost so reviewers or auditors avoid integration mistakes.

This skill enforces a mandatory, comprehensive quality loop for all code changes. It guides users through running multiple specialized validation tools—including linters, type checkers, and security scanners—for diverse ecosystems (e.g., Node.js/TypeScript, Python). It ensures that code is robust, adheres to established style guides, and maintains type safety across all components before it can be considered complete or committed.

Guides LLM post-training with reinforcement learning using the Slime (Megatron + SGLang) framework, covering GLM-style training, custom data buffers, rollout management, and Megatron-LM integration for scalable policy tuning.

Guides developers through defining scopes, creating multi-layer strategies, implementing Jest/pytest-style suites, executing tests, triaging flaky runs, and documenting coverage gaps, performance benchmarks, and security findings for QA, regression, or automation work.

Guides mechanistic interpretability research with TransformerLens by exposing HookPoints, activation caching, activation patching, and circuit-analysis workflows to inspect and manipulate GPT-style models for reverse-engineering attention behavior and causal tracing.

Guides RLHF-style LLM fine-tuning with verl's HybridFlow stack, scaling PPO/GRPO/DAPO across FSDP/Megatron/vLLM backends for math reasoning, agentic multiturn, and vision-language workflows.

Guides hypothesis testing, regression, correlation, and Bayesian analyses with automated assumption checks, diagnostics, power planning, and APA-style reporting so researchers can choose proper tests, interpret effect sizes, and document academic findings.

Guides product teams through Geoffrey Moore-style positioning statements to clarify target customer, underserved need, category, benefit, and differentiation before messaging or launch planning.

Guides teams through writing Mike Cohn‑style user stories with Gherkin acceptance criteria so product needs become actionable, testable development work with shared understanding.