Create Playground Skill

Create and maintain an interactive Streamlit playground that enables "learn by playing" instead of "learn by reading code."

Goal

Build and evolve a web-based playground where users can:

1. Explore VRP problems interactively (select, generate, visualize instances)
2. Experiment with algorithms (configure parameters, run solvers, compare results)
3. Learn through interaction (understand interfaces, pipelines, mechanisms)
4. Reproduce experiments (save configs, replay runs, export results)

Core Philosophy

Following the vision in playground/VISION.md:

• Three-layer learning: Interface → Pipeline → Mechanism
• Minimal cognitive load: Only expose what's needed for current task
• Contract-based trust: Playground behavior matches actual code (verified by tests)
• Just-in-time learning: Dive deeper only when hitting limitations