Lesson 5 of 15
Inside the harness: context, tools and permissions
Instruction files, compaction, subagents, hooks, permission modes and the sandbox, and which of them the model is free to ignore.

In this lesson
- Instructions the model reads and rules the harness enforces
- Instruction files: CLAUDE.md and AGENTS.md
- Compaction, clearing and loading on demand
- Subagents: work in a separate context window
- Hooks: rules the harness runs every time
- Permission modes, rules and the sandbox
Ultra Premium
Unlock this lesson
Ultra Premium opens the 6 premium lessons in Building with agents, along with the rest of WhiteBox's premium lessons and problems.
Not ready yet? Read the free lesson: What a coding agent actually is