dotnet-ai-toolkit

Full-project health audit — spawns 5 specialists (dotnet-reviewer, security-analyzer, testing-reviewer, performance-profiler, deployment-validator) in parallel. Use for onboarding or pre-release.

Analyze project/namespace boundaries, detect circular refs and layering violations via dotnet-depends and solution graph. Use before major refactors or to validate Clean Architecture.

Senior-engineer challenge mode — aggressively question the current approach, hunt for N+1 queries, captive deps, missing auth, sync-over-async, leaks. Use before finalizing significant changes.

Capture a solved problem as institutional knowledge. Writes `.claude/solutions/{category}/{slug}.md` with symptoms, root cause, fix, category, tags. Use after notable fixes.

Generate or update documentation — XML doc comments, README, architecture docs, OpenAPI descriptions. Uses code as source of truth; no hallucinated features.

Extract durable lessons from a just-finished fix — propose CLAUDE.md rule, Iron Law addition, or Solution doc. Use after resolving a surprising/repeatable bug.

Validate EF Core migrations for destructive ops, downtime risk, and backward compatibility. Detects dropped columns, NOT NULL additions without defaults, data-destructive renames.

Scan for EF Core N+1 query patterns — loops over materialized lists that access navigations, missing Include, lazy loading in hot paths. Reports file:line findings.

Fetch NuGet package metadata, versions, vulnerabilities, compat matrix, and deprecation status. Supports dotnet list package and NuGet.org v3 API.

Performance analysis — EF query plans, async bottlenecks, GC pressure, BenchmarkDotNet setup, allocation profiling. Spawns performance-profiler agent.

Plan a .NET feature before coding. Spawns planning-orchestrator which researches, decomposes tasks, and writes `.claude/plans/{slug}/plan.md`. Use for any non-trivial change.

Research .NET topics externally — docs.microsoft.com, NuGet registry, GitHub issues. Routes to web-researcher or nuget-researcher based on query type.

Review .NET code changes via parallel multi-track review. Spawns dotnet-reviewer, testing-reviewer, security-analyzer, iron-law-judge, verification-runner. Use after /dotnet:work or for PR review.

Scan for technical debt — TODOs, obsolete APIs, EOL frameworks, nullable warnings suppressed, analyzer warnings silenced, deprecated NuGets. Produces ranked backlog.

Convert a review verdict into prioritized work items. Reads consolidated.md, splits findings into tasks grouped by severity, produces a triage plan. Use after /dotnet:review.

Run .NET verification loop — restore → build → format → test → analyzer pass. Discovers solution/projects from .sln/global.json. Use after edits or before commit.

Execute a plan file. Parses task annotations ([direct]/[ef]/[api]/[blazor]/[test]/[security]), routes to specialists, updates checkboxes, writes progress.md. Use after /dotnet:plan.

Full-project health audit — spawns 5 specialists (dotnet-reviewer, security-analyzer, testing-reviewer, performance-profiler, deployment-validator) in parallel. Use for onboarding or pre-release.

Analyze project/namespace boundaries, detect circular refs and layering violations via dotnet-depends and solution graph. Use before major refactors or to validate Clean Architecture.

Senior-engineer challenge mode — aggressively question the current approach, hunt for N+1 queries, captive deps, missing auth, sync-over-async, leaks. Use before finalizing significant changes.

Capture a solved problem as institutional knowledge. Writes `.claude/solutions/{category}/{slug}.md` with symptoms, root cause, fix, category, tags. Use after notable fixes.

Generate or update documentation — XML doc comments, README, architecture docs, OpenAPI descriptions. Uses code as source of truth; no hallucinated features.

Extract durable lessons from a just-finished fix — propose CLAUDE.md rule, Iron Law addition, or Solution doc. Use after resolving a surprising/repeatable bug.

Validate EF Core migrations for destructive ops, downtime risk, and backward compatibility. Detects dropped columns, NOT NULL additions without defaults, data-destructive renames.

Scan for EF Core N+1 query patterns — loops over materialized lists that access navigations, missing Include, lazy loading in hot paths. Reports file:line findings.

Fetch NuGet package metadata, versions, vulnerabilities, compat matrix, and deprecation status. Supports dotnet list package and NuGet.org v3 API.

Performance analysis — EF query plans, async bottlenecks, GC pressure, BenchmarkDotNet setup, allocation profiling. Spawns performance-profiler agent.

Plan a .NET feature before coding. Spawns planning-orchestrator which researches, decomposes tasks, and writes `.claude/plans/{slug}/plan.md`. Use for any non-trivial change.

Research .NET topics externally — docs.microsoft.com, NuGet registry, GitHub issues. Routes to web-researcher or nuget-researcher based on query type.

Review .NET code changes via parallel multi-track review. Spawns dotnet-reviewer, testing-reviewer, security-analyzer, iron-law-judge, verification-runner. Use after /dotnet:work or for PR review.

Scan for technical debt — TODOs, obsolete APIs, EOL frameworks, nullable warnings suppressed, analyzer warnings silenced, deprecated NuGets. Produces ranked backlog.

Convert a review verdict into prioritized work items. Reads consolidated.md, splits findings into tasks grouped by severity, produces a triage plan. Use after /dotnet:review.

Run .NET verification loop — restore → build → format → test → analyzer pass. Discovers solution/projects from .sln/global.json. Use after edits or before commit.

Execute a plan file. Parses task annotations ([direct]/[ef]/[api]/[blazor]/[test]/[security]), routes to specialists, updates checkboxes, writes progress.md. Use after /dotnet:plan.

Self-improving loop for plugin skills. Reads program.md, proposes one mutation per iteration, evaluates against deterministic scorer, keeps improvements via git, reverts failures. Targets weakest skill+dimension. Use with /loop for overnight runs.

ASP.NET Core API patterns — Minimal APIs vs Controllers, routing, DTOs, validation, middleware, versioning, ProblemDetails. Auto-loads for API work.

Blazor Server/WASM/Auto render modes, streaming SSR, state management, forms, JS interop, lifecycle, SignalR. Auto-loads for .razor work.

Exploratory design discussion — generate and pressure-test architectural options for .NET problems before committing to a plan. Use when scope is unclear or multiple approaches exist.

Walkthrough a plan file interactively. Reads plan.md and explains decisions, task flow, and risks section by section without executing. Use before /dotnet:work on unfamiliar plans.

Schema + conventions for compound solution docs. Used by /dotnet:compound to write durable problem→solution records in .claude/solutions/{category}/.