ASP.NET Community Standup: Better AI for .NET developers with dotnet/skills
Daniel Roth and guests walk through how dotnet/skills is intended to improve AI-assisted .NET development by giving assistants structured skills/tools and validating behavior with practical evaluation scenarios.
Overview
The session focuses on improving the quality of AI-generated .NET code (accuracy and idiomatic patterns) by:
- Defining reusable skills and tools that an AI assistant can call to perform .NET-specific tasks
- Using real-world evaluation scenarios to measure whether the assistant produces correct, idiomatic results
What dotnet/skills is aiming to solve
AI assistants can generate plausible-looking code that:
- Misses .NET-idiomatic patterns
- Uses incorrect APIs or outdated approaches
- Fails on edge cases that show up in real applications
dotnet/skills is presented as a way to reduce those issues by grounding the assistant in explicit capabilities (skills/tools) and validating outcomes with scenario-based evaluation.
Key themes discussed
- More accurate .NET code generation: improving correctness and reducing hallucinated or mismatched APIs
- Idiomatic .NET output: steering assistants toward patterns that .NET developers expect
- Skills/tools approach: packaging repeatable capabilities that can be invoked by an assistant
- Evaluation scenarios: testing assistant behavior against realistic tasks to see what works and what needs improvement
People featured
- Daniel Roth
- Mike Kistler
- Javier Calvarro Nelson
- Wendy Breiding
- Abhitej John Bandi