Explanation

Conceptual deep-dives into how proto works. Read these to build a mental model of the system.

• Architecture — How the CLI, core, and SDK fit together
• Agent Harness — The loop that drives multi-step agentic behavior
• Sub-Agents — How sub-agents are spawned, isolated, and composed
• Skills — How skills are discovered, loaded, and injected into context
• Hooks — The event system and how hooks interact with the agent loop
• Approval Modes — How tool permission decisions are made