AGENTS.md — Guidance for Codex Agents

Project: Wanderland
Purpose: A radically simplified land finder that overlays and filters pre-made GeoJSONs to help users find sovereign land in Canada. No ETL. No data science. Just maps.

Project Overview

Wanderland is a GeoJSON-first mapping tool:

• All datasets are stored as GeoJSON
• The app loads, overlays, and filters these datasets in real time
• The user experience is an interactive map with filter controls

Agent Goals

Codex Agents should:

• Prioritize clarity over cleverness
• Write modular filters (one function = one filter)
• Always assume input and output are GeoJSON
• Never introduce unnecessary data processing pipelines
• Keep functions pure and auditable
• Make sure to document changes and update them if needed in the readme and in accompanying /docs folder.

Code Conventions

Languages:

• JavaScript/TypeScript (frontend)
• Python (optional for scoring or batch jobs)

Standards:

• JS/TS: prettier, eslint
• Python: black, flake8
• Filenames: snake_case.py, camelCase.js

Style:

• Use named imports/exports
• Comment all public functions and complex logic
• Keep files small; one feature = one file

Filters: Key Rules

• Filters operate on GeoJSON FeatureCollections
• Filters can be attribute-based (e.g., property thresholds) or geometry-based (e.g., distance to water)
• Filters should be composable:

const nearWater = (feature) => feature.properties.water_distance < 1000;
const lowTax = (feature) => feature.properties.tax_rate < 0.05;

• The UI applies one or more filters, producing a filtered GeoJSON layer

PR Guidelines

Title:

[Filter/Feature/Fix] Short description

Body:

• What changed and why
• How to test it
• Screenshots (if applicable)

Minimum Working Product (MVP)

Must support:

• Loading multiple GeoJSON layers
• Applying composable filters
• Displaying filtered map overlays
• Optional: exporting filtered results

Do's & Don'ts

✅ DO:

• Use GeoJSON everywhere
• Build tiny, reusable filter functions
• Keep the interface dead simple

❌ DON'T:

• Add complex data processing or server-side transformations
• Use non-spatial formats unless absolutely required
• Build monolithic code blocks

Final Note

Wanderland is for visual decision-making, not for perfect data science. The priority is speed, clarity, and sovereignty-first thinking.

Build tools that help people see, not tools that bury them in complexity.