AI Prompts Overview

This application was developed using Claude Code (Anthropic's CLI tool for Claude). This section documents all prompts used during development for transparency and reproducibility.

What is Claude Code?

Claude Code is an AI-powered coding assistant that can:

• Read and write files
• Execute shell commands
• Search codebases
• Create and modify code
• Generate documentation

Development Session

The M-580 2026 map was created in an interactive session with the following major phases:

Phase 1: Initial Map Development

Creating the base MapLibre GL JS application with:

• Interactive map with federal land boundaries
• Sidebar with searchable unit list
• Agency filtering
• Popup information displays

View Map Development Prompts

Phase 2: Data Processing

Processing and validating data including:

• Extracting student home units from PDF
• Verifying GeoJSON data coverage
• Adding missing units and offices

View Data Processing Prompts

Phase 3: Leaflet Fallback

Creating a WebGL-free version for VM compatibility:

• Converting MapLibre code to Leaflet
• Implementing Canvas rendering
• Adjusting layer ordering

View Map Development Prompts

Phase 4: Documentation

Creating MkDocs documentation:

• User guides
• Development setup
• Prompt documentation

View Documentation Prompts

Prompt Engineering Principles

The following principles were used when crafting prompts:

1. Be Specific

❌ "Make a map"
✓ "Create a MapLibre GL JS map showing federal land units with a sidebar,
   search functionality, and agency filters"

2. Provide Context

❌ "Fix the error"
✓ "WebGL context creation is failing because I'm on a remote VM without
   GPU support. Create a Leaflet version that uses Canvas rendering instead."

3. Reference Existing Files

❌ "Check if the data is complete"
✓ "Compare the student list in /home/user/Downloads/M580-2026-Selections.pdf
   against the units in data/units.geojson and data/offices.geojson"

4. Iterate and Refine

Development proceeded through iterative prompts:

1. Create initial implementation
2. Test and identify issues
3. Request specific fixes
4. Add enhancements

Model Information

• Model: Claude Opus 4.5 (claude-opus-4-5-20251101)
• Interface: Claude Code CLI
• Session Date: January 2026

Reproducibility

To reproduce this application:

1. Install Claude Code
2. Start a new session in an empty directory
3. Follow the prompts documented in this section
4. Adjust paths and filenames as needed