The AGENTS.md File Changes Everything

Key Takeaways

• AGENTS.md defines project conventions for AI code generation
• Vertical slice architecture groups code by feature, not layers
• AI skills in .claude/skills streamline prompts and reduce token cost
• IDbContextFactory resolves scoped DbContext bugs in Blazor Server
• Workshop shows how to build a Blazor app from scratch in 30 minutes

Pulse Analysis

AI‑assisted development is reshaping how .NET teams bootstrap projects. By placing a concise AGENTS.md file at the repository root, developers give large language models a permanent style guide that enforces vertical‑slice architecture, dependency‑injection standards, and naming conventions. This eliminates the repetitive scaffolding that traditionally consumes hours, allowing engineers to focus on business logic while the AI handles boilerplate. The approach also creates a single source of truth for code quality, making onboarding and code reviews more efficient.

The real power emerges when the AGENTS.md rules are broken into modular skills under a .claude/skills folder. Each skill acts as a targeted checklist—one for backend vertical slices, another for Blazor UI components—so prompts stay short and token‑efficient. When a developer asks Claude to "create a vertical slice for X," the model pulls the appropriate skill, adheres to the predefined folder layout, and outputs files that fit the project's architecture without extra prompting. This modularity scales across teams and reduces the risk of divergent code structures.

Practical experience shows the method’s limits and benefits. In the showcased BookRatings app, AI quickly scaffolded entities, EF Core in‑memory data stores, and UI pages, but it also introduced a needless repository layer and a scoped DbContext issue. The author resolved the latter by switching to IDbContextFactory, demonstrating that developers must still validate AI output. The blend of automated speed and human oversight creates a sustainable workflow, and the upcoming two‑hour live Blazor AI workshop promises to teach participants how to replicate the process from a blank folder to a deployable Azure solution.