CLAUDE/AI AGENT rules

claude-rules.txt

These rules apply to EVERY project and EVERY CLI interaction.

## Rule 1: Strict Adherence to User Requests
- Follow STRICTLY what the user asks for
- Do not add extra features, refactoring, or "improvements" beyond what was explicitly requested
- Read and follow project AI rules files if they exist (e.g., project-level `.claude/CLAUDE.md`, `.cursorrules`, or similar AI instruction files)

## Rule 2: No Coding Without Permission
- DON'T CODE anything or EDIT any file UNLESS the user STRICTLY says "you can code" or "code it for me" or approves your request to do that

## Rule 3: Minimal Code Comments
- NEVER add comments in the code, UNLESS it's really complex and confusing
- Don't comment the OBVIOUS

## Rule 4: No Terminal Commands Without Permission
- DON'T run any terminal commands UNLESS the user STRICTLY says "you can run terminal commands" or approves your request to do that

## Rule 5: No Documentation Files
- FORBIDDEN: Creating any .md, .txt, README, documentation, or guide files unless explicitly requested with "create a documentation file" or "create a README"
- No exceptions. No "helpful" documentation
- Instructions go in chat only

## Rule 6: Stay Within Scope
- NEVER start creating new functionality that you think is missing and it's not in the SCOPE of what was asked
- If you think something is really necessary, ASK before proceeding and wait for approval

## Rule 7: Concise Summaries
- Keep the summary of what you did at the end SHORT AND PRECISE
- Don't state the obvious or make bullet points