Skip to content

/roslyn-query

Query .NET/C# codebases using Roslyn AST analysis via dotnet run file. Use for tracing data flow, auditing API usage, finding pattern violations, or ad-hoc codebase queries.

shell
$ npx -y skills add NikiforovAll/claude-code-rules --skill roslyn-query --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/roslyn-query
How auto-invocation works

Context preview

The summary Claude sees to decide when to auto-load this skill.

Query .NET/C# codebases using Roslyn AST analysis via dotnet run file. Use for tracing data flow, auditing API usage, finding pattern violations, or ad-hoc codebase queries.

SKILL.md

roslyn-query.SKILL.md
name: roslyn-query
description: Query .NET/C# codebases using Roslyn AST analysis via dotnet run file. Use for tracing data flow, auditing API usage, finding pattern violations, or ad-hoc codebase queries.
argument-hint: "[analysis description or query type]"

Roslyn Query

Query .NET/C# codebases using Roslyn AST analysis scripts executed via `dotnet run file.cs`.

Before generating scripts, load the `handbook-dotnet:dotnet-run-file` skill for execution conventions and directive syntax.

When to Use

  • Trace data flow from sources to sinks (taint analysis)
  • Audit all assignments to specific properties
  • Find all callers of a method
  • Find all implementations of an interface
  • Detect pattern violations (empty catches, missing sanitizers, etc.)
  • Architecture validation (layer violations)
  • Ad-hoc codebase structural queries

Analysis Tiers

Tier 1: Syntax-Only (fast, no compilation)

Uses `CSharpSyntaxTree.ParseText()` — parses files individually, no solution loading.

Best for:

  • Property assignment audits (`x.Content = ...`)
  • Empty catch block detection
  • Naming convention checks
  • Cyclomatic complexity
  • Magic number/string detection
  • Method size metrics
  • Pattern matching on code structure

Tier 2: Semantic (needs project/solution)

Uses `MSBuildWorkspace` — loads project, resolves types, enables cross-file analysis.

Best for:

  • Find all callers of a method (`SymbolFinder.FindCallersAsync`)
  • Find interface implementations (`SymbolFinder.FindImplementationsAsync`)
  • Find method overrides (`SymbolFinder.FindOverridesAsync`)
  • Type dependency graphs
  • Full data flow analysis (`SemanticModel.AnalyzeDataFlow`)

Instructions

Step 1: Determine analysis type

Based on the user's request, choose the appropriate template from `references/templates.md`. If no template fits, compose a custom script using the patterns documented there.

Step 2: Generate the script

Create a `.cs` file in a `scripts/` directory (or temp location) using `dotnet run file.cs` format with `#:package` directives.

For syntax-only analysis:

#:package Microsoft.CodeAnalysis.CSharp@4.*

For semantic analysis, additionally:

#:package Microsoft.CodeAnalysis.CSharp.Workspaces@4.*
#:package Microsoft.CodeAnalysis.Workspaces.MSBuild@4.*
#:package Microsoft.Build.Locator@1.*
#:property DisableMSBuildAssemblyCopyCheck=true

> **Note:** `DisableMSBuildAssemblyCopyCheck=true` suppresses the MSBL001 build error that requires `ExcludeAssets="runtime"` on MSBuild packages — an attribute `#:package` directives can't express. Safe for throwaway analysis scripts. > > **`.slnx` not supported:** Roslyn 4.x `MSBuildWorkspace` only supports `.sln`, not `.slnx`. Use `OpenProjectAsync` with a `.csproj` instead — transitive project references are loaded automatically.

Step 3: Script conventions

  • Accept target directory/path via `args[0]`, default to `Directory.GetCurrentDirectory()`
  • Use `Path.GetRelativePath()` for output paths
  • Print structured output: file:line, description, context
  • Use color: green for safe, yellow for warnings, red for violations
  • Exit code 0 = clean, 1 = findings

Step 4: Execute

dotnet run scripts/my-analysis.cs -- <target-path>

Step 5: Interpret results

Analyze the output. For taint analysis, categorize findings:

  • **Safe** — trusted source (code-generated, AI output, literal null)
  • **Already sanitized** — sanitizer call deeper in call chain (syntax-only false positive)
  • **Needs investigation** — external input, user-authored content
  • **Violation** — clear miss that should be fixed

Key Roslyn APIs Quick Reference

Load `references/roslyn-api.md` for detailed API reference with code examples.

Templates

Load `references/templates.md` for ready-to-use analysis script templates.

Read more
Read it on GitHub ↗
Ships withclaude-code-rules

A collection of Claude Code recommendations and practices. Learn practical techniques to enhance your AI-assisted development workflow with Claude Code.

Get the whole plugin, auto-invoked
Stats
136
Stars
0
Views
19
Forks
Active
Maintenance
HTML
Language
Apache-2.0
License
5d ago
Last commit
1y ago
Created

Repo: NikiforovAll/claude-code-rules

Other skills on claude-code-rules.