sherlock

Sherlock: .NET assembly introspection

Sherlock reads real assemblies (your build output or the NuGet cache) via metadata-only reflection, so its answers match the exact version the project compiles against. Reach for it whenever correctness depends on an API's actual shape.

Tool names are snake_case; argument names are camelCase.

1. Locate the assembly

Every tool takes an assembly path. Find it rather than guessing:

• find_assembly_by_class_name - you know a type name but not which DLL holds it.
• find_assembly_by_nuget_package - resolve a package (and version) from the local NuGet cache.
• get_project_output_paths - the build output DLL of a .csproj.
• find_assembly_by_file_name - you know the DLL name.

Never hardcode bin/Debug/<tfm>/...; the target framework and configuration vary.

When you will make several calls against the same assembly, call open_assembly once and pass the returned assemblyHandle (a short asm_… id) instead of assemblyPath. The handle also carries any additionalAssemblies you opened it with.

2. Orient cheaply

• get_assembly_info - identity, target framework, references.
• search_members - you know a member name fragment but not its declaring type (narrow with memberKinds).
• get_types_from_assembly - browse types, paginated.

3. Drill in, narrow to wide

1. get_type_info - kind, base type, interfaces, accessibility, member counts.
2. get_type_members filtered by kinds, nameContains or hasAttributeContains. The default summary projection returns { kind, name, signature }, and the C# signature already carries return type, parameters and modifiers.
3. Re-call with projection='full' only for the few members where you need structured parameters, attributes or modifier flags.
4. analyze_method for one method's overloads; get_xml_docs_for_type / get_xml_docs_for_member for documentation; get_type_hierarchy for the inheritance chain (format='mermaid' for a class diagram).

Start with a small maxItems and page with continuationToken instead of fetching everything.