Guidelines for AI / Codex Edits¶

This page defines explicit rules for using AI tools (such as GitHub Copilot, ChatGPT, or Codex) when editing this repository. The goal is to maintain scientific accuracy, structural integrity, and documentation consistency.

Core rules¶

1. Do NOT invent scientific claims¶

All scientific statements must come from: - NEON documentation
- sensor metadata
- existing package logic
- peer-reviewed sources

2. Do NOT alter conceptual meaning without review¶

Edits must preserve: - scientific correctness
- processing assumptions
- pipeline logic

Documentation layout is intentional and should not be reorganized automatically.

4. Do NOT remove placeholders, schemas, or metadata¶

These are required for reproducibility.

Code-editing rules¶

Do not introduce new APIs without discussion
Do not refactor internal architecture without review
Avoid automatic renaming of variables or functions
Confirm that unit tests pass after any edit

Documentation-editing rules¶

Only modify the files explicitly referenced in a prompt
Never rewrite entire sections unless provided with exact text
Preserve all fenced code blocks
Maintain consistency with mkdocs navigation

When NOT to use AI tools¶

Avoid automatic editing when:

adding new scientific explanations
modifying BRDF/topographic formulas
updating spectral response files
changing reflectance scaling logic

These require domain expertise.

Safe tasks for AI tools¶

inserting text provided by the user
reorganizing text when explicitly instructed
generating placeholder scaffolding
performing mechanical refactors with tests

Summary¶

AI tools in this repository should behave like careful editors—not authors.
Always prioritize scientific correctness, reproducibility, and transparency.