SKILL.md
Project Setup — Tech-Stack Advisor
This skill owns one thing: the kit's recommended tech-stack defaults and why. The workflows that consume it live elsewhere:
- Initializing a project / generating CLAUDE.md →
dotnet-init(interactive flow, architecture questionnaire, CLAUDE.md generation) - Assessing an existing codebase →
health-check(the canonical 8-dimension graded assessment) - EF Core schema, NuGet, or .NET version migrations →
migrate - Choosing an architecture →
architecture-advisor(always ask before recommending)
Core Principles
- Recommend a default, explain the why, let the user choose — Every dimension has a kit default, but defaults are starting points, not mandates. State the trade-off in one line so the choice is informed.
- Prefer built-in .NET over third-party —
HybridCacheover Redis-client wrappers, built-in rate limiting over packages, built-in OpenAPI over Swashbuckle. Fewer dependencies means fewer licensing surprises and upgrade breaks. - License-aware picks — MediatR (v13+), MassTransit (v9+), and FluentAssertions (v8+) went commercial. The kit defaults to MIT alternatives: Mediator, Wolverine, plain xUnit asserts.
- Add messaging later, not never — Most projects don't need a message bus on day one. Default to "None (add later)" and reach for Wolverine when async workflows actually appear.
Patterns
Tech-Stack Dimensions and Defaults
| Dimension | Options | Default | Why |
|---|---|---|---|
| Database | PostgreSQL, SQL Server, SQLite | PostgreSQL | Open source, best EF Core provider outside SQL Server, first-class Testcontainers support |
| Auth | JWT Bearer, OIDC (Keycloak/Auth0), None | JWT Bearer | Simplest secure default for APIs; move to OIDC when an external IdP exists |
| Caching | HybridCache, Redis, None | HybridCache | Built-in, stampede protection, L1+L2 — add Redis only as its L2 backend |
| Messaging | Wolverine (RabbitMQ), MassTransit, None | None (add later) | Premature messaging adds ops burden; Wolverine (MIT) when needed |
| Observability | Serilog + OpenTelemetry, Basic logging | Serilog + OTEL | Structured logs + traces from day one are cheap; retrofitting is not |
| Resilience | Polly v8 pipelines, Basic retry | Polly v8 | AddStandardResilienceHandler() is one line for production-grade defaults |
| API docs | Built-in OpenAPI + Scalar | OpenAPI + Scalar | Framework-maintained spec generation; Scalar replaces Swagger UI |
| Testing | xUnit v3 + Testcontainers | xUnit v3 + Testcontainers | Real databases in tests; in-memory providers hide real bugs |
Once dimensions are chosen, dotnet-init bakes them into the generated CLAUDE.md, and each choice maps to a skill to load when working in that area (ef-core, authentication, caching, messaging, serilog, opentelemetry, resilience, openapi, scalar, testing).
Anti-patterns
Prescribing a Stack Without Asking
# BAD — assuming the kit defaults apply everywhere
"You should use PostgreSQL and Wolverine."
# The team runs SQL Server enterprise-wide and has zero async workflows.
# GOOD — default + trade-off + question
"Kit default is PostgreSQL (best OSS EF provider). Any organizational
constraint — existing SQL Server licenses, DBA support — that should
override it?"Re-Running Workflows This Skill Doesn't Own
# BAD — improvising a health grading or init flow from this skill
"Let me grade your codebase across 5 categories..."
# That grading conflicts with the canonical one.
# GOOD — route to the owner
Init/CLAUDE.md → dotnet-init | Assessment → health-check | Upgrades → migrateDecision Guide
| Scenario | Route to |
|---|---|
| "Set up this project for Claude Code" | dotnet-init |
| "Which database/auth/caching should I use?" | This skill — table above |
| "How healthy is this codebase?" | health-check |
| "Upgrade to .NET 10" / "update packages" | migrate |
| "Which architecture fits?" | architecture-advisor |
| Stack chosen, ready to build | scaffold for the first feature |