Tag

#claude

AI: global rules

Global AI rules

The rules

# Global Rules
1. **Project context** — Read AGENTS.md before working in a repo
   CLAUDE.md, GROK.md and other agent-specific files must direct agents to AGENTS.md
   and contain only agent-specific setup
2. **Scope** — Do what was asked. If the task needs to go wider, explain and ask.
3. **Reuse** — Prefer existing code
4. **Conventions** — Match the surrounding code: naming, structure, error handling.
   Check a neighbouring file before inventing a pattern
5. **Reporting** — State what you actually ran and summarize the results.
   Distinguish verified results from assumptions
6. **Hard no** — Never print, log or commit secrets or .env contents.
7. **Git** — Don't push unless asked
8. **Graphify** — Use only in software projects, for architecture, cross-file
   relationships or locating functionality
9. **Uncertainty** — Do not invent facts, sources or results. If you don't know,
   say so

Location

Agent Where
Claude Desktop / Claude.ai Settings → General → Instructions for Claude
Claude Code ~/.claude/CLAUDE.md
Codex ~/.codex/AGENTS.md
Grok ~/.grok/AGENTS.md

← All tags