The AI Coding Dictionary describes various operational presets for agents, known as permission or approval modes, which dictate their behavior during runtime, such as blocking edits or auto-approving actions. These modes can be switched mid-session to adapt to different tasks or requirements.
The concept of "permission modes" or "approval modes" that dictate AI agent behavior at runtime offers a fresh content angle. You could explore how these modes impact AI productivity and security, comparing terms like Claude Code's "permission modes" and Codex's "approval modes," while discussing their evolution to include behavioral bundling. This provides a unique narrative on managing AI agents for different coding and research needs, such as "plan mode" for research or "YOLO mode" for unrestricted actions.