Designing AI Coding Tools: Inside Visual Studio & VS Code Agents
Rohan Malpani (Principal Designer for Visual Studio & VS Code) discusses how AI is reshaping developer tools, with a focus on practical workflows teams use to ship production-ready, AI-assisted code.
Overview
Why designers should help build coding tools
The episode frames AI-assisted development as a collaboration problem as much as a tooling problem: designers and developers need shared workflows and shared language to move from prototypes to features that can ship.
Visual Studio “skills”: reusable building blocks
Rohan explains Visual Studio “skills” as reusable, repeatable units of capability—described as working like reusable “recipes”.
Key ideas covered:
- Skills are intended to be reused across scenarios rather than being one-off prompts.
- The workflow emphasizes repeatability and consistency so teams can rely on outcomes.
Creating a skill with AI (live demo)
The discussion includes a live walkthrough of creating a skill, focusing on how the experience is designed so that building and reusing a skill fits into day-to-day development.
VS Code Agents Window
Rohan gives an overview of the Agents Window experience in VS Code and how it supports agent-driven workflows for building and iterating on developer tools.
Agents vs. skills
The episode contrasts the two concepts:
- Skills: positioned as reusable “recipes” that can be applied repeatedly.
- Agents: positioned as a workflow for iterating on work inside a development environment, including building features and improving tooling experiences.
Using agents to build developer tools in a real repo
A recurring theme is working “inside your own repo” and iterating on features in a way that aligns with real engineering constraints.
Topics highlighted:
- Designing with AI in the context of an existing codebase.
- Iterating on features with agents as part of an engineering workflow.
Moving beyond “vibe coding” to production workflows
The conversation explicitly calls out the gap between quick prototypes and production engineering, and focuses on what teams do to close that gap.
Code quality and security considerations
The episode discusses how teams ensure quality and security when using AI-assisted coding, emphasizing that shipping requires guardrails and review practices rather than relying on ad-hoc generation.
Advice for designers and developers
Rohan closes with guidance for teams adopting AI in their workflows, focusing on collaboration between design and engineering and on building repeatable practices that scale beyond demos.