How to use the Backend Dev Guidelines agent skill

What the skill makes the agent do

An agent skill is a folder of instructions that an AI agent loads when a task matches the skill's description. This one covers routes, controllers, services, repositories, middleware, Prisma database access, Sentry error tracking, Zod validation, unifiedConfig and async patterns. When it applies, the agent follows a layered design: routes → controllers → services → repositories → database.

The skill is reference material only. It is markdown guides and code examples, and it does not run anything itself.

• Keeps route files thin and delegates to controllers.
• Has controllers extend BaseController.
• Validates input with Zod schemas.
• Reads settings from unifiedConfig instead of process.env.
• Reports errors to Sentry with Sentry.captureException.
• Uses a repository pattern for data access instead of calling Prisma everywhere.

Install the backend-dev-guidelines skill

Download the ZIP file from this page and unzip it. The skill folder should contain SKILL.md and the resources folder with its markdown files.

For a personal setup, unzip it into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder instead.

In the Claude apps, you can also upload the skill under Settings → Capabilities → Skills.

Trigger the skill with natural requests

The skill activates automatically when your request involves the topics in its description. You do not need to name it, but mentioning the stack helps. The source lists triggers such as creating or modifying routes, endpoints and APIs, building controllers, services and repositories, implementing middleware, Prisma database work, Sentry error tracking, Zod validation, configuration and backend testing.

• Add a POST endpoint for creating invoices, with Zod validation and a service layer.
• Move the logic out of this Express route file into a controller and service.
• Add Sentry error capture to the payment repository calls.
• Refactor this legacy Express handler toward the layered architecture.
• Write unit and integration tests for the UserService.

Know what the skill needs in your project

The skill is written for a Node.js, Express and TypeScript project. Its examples assume Prisma, Sentry and Zod are installed. The examples also reference your repo's own unifiedConfig and BaseController modules, which the skill does not include. If your repo does not have them, the agent will be building against modules that do not exist unless you create equivalents.

The source says that for a different stack you should adapt the architectural principles rather than copy the examples. It points to CLAUDE_INTEGRATION_GUIDE.md in the repo root for guidance on that.

• Node.js, Express and TypeScript project.
• Prisma, Sentry and Zod installed.
• Your own unifiedConfig and BaseController modules.

Review the code the agent writes

Treat the output as a draft that follows house rules, and check it against your codebase. The skill's own checklists include unit and integration tests, a Sentry entry for error tracking and a unifiedConfig setup. The anti-patterns list names direct process.env use, missing error handling, missing input validation and console.log in place of Sentry.

Because the skill says nothing about running scripts or using the network, check any commands the agent proposes before running them.

• Confirm that imports point to modules that exist in your repo.
• Confirm that errors are captured and rethrown, not swallowed.
• Run your own tests after the agent makes changes.