Skill UI

This skill provides structured frameworks for navigating difficult workplace conversations and delivering effective feedback. It covers the Preparation-Delivery-Follow-up model and Situation-Behavior-Impact (SBI) technique to help manage conflicts, address performance issues, and align expectations during 1:1 meetings.

Use the Figma MCP server to fetch design context, screenshots, variables, and assets from Figma nodes, then translate them into production code that follows your project's design system. Covers setup, troubleshooting, and link-based prompting.

This skill enables developers to translate Figma design nodes into production-ready code with pixel-perfect accuracy using the Figma MCP workflow. It ensures consistent integration with design tokens and visual parity. The process involves fetching design context, capturing screenshots, downloading assets, and adapting the design to project conventions. It requires a connected Figma MCP server and specific node IDs or desktop selections.

Convert local image, audio, video, document, and data files using installed tools while preserving originals and verifying output.

Provides methodologies to identify, exploit, and remediate file path traversal and local file inclusion vulnerabilities in web applications.

Handles file uploads and cloud storage with S3 and Cloudflare R2, covering presigned URLs, multipart uploads, image optimization, large-file handling, and upload security.

Systematically review local branch changes for bugs, security vulnerabilities, and code quality issues using a structured five-phase audit covering injection, XSS, auth, CSRF, race conditions, and more.

Expert guidance for fine-tuning LLMs with Axolotl, covering YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, and multimodal training workflows.

This skill provides expert guidance for fine-tuning Large Language Models using the LLaMA-Factory framework. It supports WebUI no-code operations, covers over 100 models, and offers various quantization options like QLoRA. Users can debug code, implement solutions, and learn best practices through comprehensive documentation and reference files.

Fine-tune large language models by training under 1% of parameters using LoRA, QLoRA, and 25+ PEFT adapter methods, enabling memory-efficient training and multi-adapter serving.

Expert guidance for fast LLM fine-tuning with Unsloth, covering LoRA/QLoRA optimization, 2-5x faster training, 50-80% memory savings, and best practices for Llama, Mistral, Gemma, and Qwen models.

Routes fine-tuning decisions: decide whether to fine-tune, choose SFT, DPO/ORPO/KTO, GRPO/RLVR, continued pretraining, and base model size. Helps avoid unnecessary training and select the right method.