Skill UI

Reduces token costs by scoping context into directory-level CLAUDE.md files. Sets up hierarchical memory, tracks savings via dashboard, routes agents to sub-contexts, and provides audit and insights commands.

Train or fine-tune language and vision models using TRL or Unsloth on Hugging Face Jobs infrastructure. Supports SFT, DPO, GRPO methods and GGUF conversion for local deployment. Requires Pro account and HF_TOKEN for Hub saving. Optimized for cloud GPU training without local setup.

This skill automatically routes subtasks to the most cost-effective Claude model (Haiku, Sonnet, or Opus) based on complexity. It logs real token usage and cost per session, and includes an A/B testing feature to validate model tier savings against your actual workload, ensuring accurate cost control.

Prove data/AI workflows end-to-end on tiny samples, find cost-saving wins, then scale within budget using cost-gated steps. Ideal for multi-stage pipelines involving paid compute like LLM tokens or warehouse credits.