Ta treść nie jest jeszcze dostępna w Twoim języku.
AI coding assistants typically offer different modes of interaction, each tailored to a specific phase of the development process. Understanding the distinction between these modes is crucial for working efficiently and safely. The two most fundamental modes are Ask (for exploration) and Agent (for execution).
Mastering when to use each mode is a key best practice that will significantly improve the quality of your AI-assisted work.
Think of the two modes as different stances you take when collaborating with your AI partner.
Ask Mode: The Explorer
Purpose: Learning, planning, and questioning.
In Ask mode, the AI is a read-only partner. It can search your entire codebase, read files, analyze dependencies, and answer your questions, but it will never make any changes to your code. It is a completely safe environment for exploration.
Agent Mode: The Implementer
Purpose: Execution, refactoring, and action.
In Agent mode, the AI has full agency. It is a read-write partner that can autonomously explore your project, edit multiple files, run terminal commands, and fix errors to carry out your instructions. This is the mode for getting work done.
The most effective and reliable workflow is to begin in Ask mode to build a solid foundation of understanding and planning, and only then switch to Agent mode for execution.
Start in Ask Mode: Understand and Plan. You’ve been tasked with refactoring a complex, unfamiliar module. Before making any changes, you start in Ask mode.
LegacyDataProcessor class? What are its main responsibilities?”processData method is called.”Switch to Agent Mode: Execute the Plan. After reviewing and refining the plan generated in Ask mode, you’re confident in the approach. Now, you switch to Agent mode to carry out the work.
processData method to be an async function.”