Agent customization
AI models contain broad general knowledge but don't know your specific codebase or team practices. Agent customization is how you share that context to make responses match your coding standards, project structure, and workflows.
This article is the decision matrix for customization: it explains the different options and helps you choose which one fits your goal. To learn what each option is and how to get started, see Customize agent behavior in Visual Studio Code and the individual guides linked from each option.
Customization options at a glance
Each option shapes a different part of how the agent works. Find the goal that matches your need, then follow the link to set it up.
| Goal | Use | Example | When it activates |
|---|---|---|---|
| Apply the same coding standards to all code | Always-on instructions | Enforce ESLint rules, require JSDoc comments | Automatically included in every request |
| Apply different rules to different file types | File-based instructions | React patterns for .tsx files |
When files match a pattern or description |
| Automate a multi-step workflow that needs scripts | Agent skills | Scaffold a service from bundled template files and a setup script | When the task matches the skill description |
| Give the AI a focused role with limited tools | Custom agents | Security reviewer, database admin | When you select it or another agent delegates to it |
| Connect the AI to external APIs or databases | MCP | Query a PostgreSQL database | When the task matches a tool description |
| Run a command automatically during the agent's work | Hooks | Run a formatter after every file edit | When the agent reaches a matching lifecycle event |
| Run a repeatable task on demand | Prompt files | Scaffold a React component | When you invoke a slash command |
Start with custom instructions for project-wide standards. Add agent skills to automate repeatable, multi-step tasks. Use MCP when you need external data. Create custom agents for specialized roles. You can combine multiple customization types as your needs grow.
Agent plugins bundle several of these options into a single installable package, letting you adopt a ready-made workflow without building it yourself.
How customizations combine
The customization options are designed to layer:
- Instructions shape how the AI writes code (conventions, style, libraries).
- Prompt files and agent skills encapsulate what the AI does for recurring tasks, from a single prompt up to a multi-step workflow with scripts.
- Custom agents define who the AI acts as (persona, tools, model), and can delegate to other agents for multi-step workflows.
- MCP servers extend what the AI can reach by adding tools that connect to external systems.
- Hooks enforce deterministic actions at specific lifecycle points in the agent loop, regardless of what the model decides to do.
- Agent plugins are pre-packaged bundles of the above, distributed through plugin marketplaces.
For configuration steps and examples, see Customize agent behavior in VS Code and the individual articles linked from the table above.