---
published: true
id: instructions
title: Instructions
description: Knowledge captured in João Lopes' learning graph.
tags:
- claude-code
- ai-engineering
locale: en
resource: https://joaolopes.dev.br/knowledge/concepts/en/instructions/
created: '2026-03-30'
---

# Instructions
What Anthropic calls CLAUDE.md files: plain markdown files loaded into the context window at session start to give the model persistent context.

## Definition
Instructions are CLAUDE.md files that load into the context window at session start. The model's response is conditioned on them, but they are not enforced. Instructions are distinct from rules: rules are path-scoped files in .claude/rules/ that load based on glob patterns, while instructions are broad and always-on. Both end up in the context window, but instructions load by directory hierarchy while rules load by file path matching. Anthropic's docs are specific: CLAUDE.md files are instructions, .claude/rules/ files are rules.

## Source
- [Manage Claude's memory - Claude Code Docs](https://docs.anthropic.com/en/docs/claude-code/memory)

## Related
- [CLAUDE.md](/concepts/en/claude-md.md)
- behavioral-modeling
- [Hooks](/concepts/en/hooks.md)