mcpbeat

Writing Server Code

bitwarden/writing-server-code

Bitwarden server code conventions for C# and .NET. Use when working in the server repo, creating commands, queries, services, or API endpoints. Also use when writing xUnit tests with `SutProvider`/`BitAutoData`, registering DI, or generating entity IDs.

1k tokens
context cost
the whole folder, loaded on every use
1
files
instructions only
0
copies elsewhere
how many repositories repackaged it
19741
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/bitwarden/server --skill writing-server-code

The instruction itself

11 sections, as written by the author

Architectural Rationale

Command Query Separation (CQS)

New features should use the CQS pattern — discrete action classes instead of large entity-focused services. See ADR-0008.

Why CQS matters at Bitwarden: The codebase historically grew around entity-focused services (e.g., CipherService) that accumulated hundreds of methods. CQS breaks these into single-responsibility classes (CreateCipherCommand, GetOrganizationApiKeyQuery), making code easier to test, reason about, and modify without unintended side effects.

Commands = write operations. Change state, may return result. Named after the action: RotateOrganizationApiKeyCommand.

Queries = read operations. Return data, never change state.

When NOT to use CQS: When modifying existing service-based code, follow the patterns already in the file. Don't refactor to CQS unless explicitly asked. If asked to refactor, apply the pattern only to the scope requested.

Caching

When caching is needed, follow the conventions in CACHING.md. Use IFusionCache instead of IDistributedCache.

Don't implement caching unless requested. If a user describes a performance problem where caching might help, suggest it — but don't implement without confirmation.

GUID Generation

Always use CoreHelpers.GenerateComb() for entity IDs — never Guid.NewGuid(). Sequential COMBs prevent SQL Server index fragmentation that random GUIDs cause on clustered indexes, which is critical for Bitwarden's database performance at scale.

Library shape

When creating or modifying code under src/Libraries/, read src/Libraries/LIBRARY.md — it is the canonical shape and covers public surface, settings, endpoints, repositories, and cross-library dependencies.

Critical Rules

These are the most frequently violated conventions. Claude cannot fetch the linked docs at runtime, so these are inlined here:

  • Use TryAdd* for DI registration (TryAddScoped, TryAddTransient) — prevents duplicate registrations when multiple modules register the same service
  • File-scoped namespacesnamespace Bit.Core.Vault; not namespace Bit.Core.Vault { ... }
  • Nullable reference types are enabled (ADR-0024) — use ! (null-forgiving) when you know a value isn't null; use required modifier for properties that must be set during construction
  • Async suffix on all async methodsCreateAsync, not Create, when the method returns Task
  • Controller actions return ActionResult<T> — not IActionResult or bare T
  • Testing with xUnit — use [Theory, BitAutoData] (not [AutoData]), SutProvider<T> for automatic SUT wiring, and Substitute.For<T>() from NSubstitute for mocking

Examples

GUID generation

// CORRECT — sequential COMB prevents index fragmentation
var id = CoreHelpers.GenerateComb();

// WRONG — random GUIDs fragment clustered indexes
var id = Guid.NewGuid();

DI registration

// CORRECT — idempotent, won't duplicate
services.TryAddScoped<ICipherService, CipherService>();

// WRONG — silently duplicates registration, last-wins causes subtle bugs
services.AddScoped<ICipherService, CipherService>();

Namespace style

// CORRECT — file-scoped
namespace Bit.Core.Vault.Commands;

// WRONG — block-scoped
namespace Bit.Core.Vault.Commands
{
    // ...
}

Further Reading

How to use it

Copy the folder

Take bitwarden/writing-server-code from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.