Score in a Platform Engineering Workflow

Where Score Fits in the Bigger Picture

Score does not replace Kubernetes, Helm, Terraform, or Docker Compose. It sits above them as an abstraction layer. Think of it as the interface between the developer experience layer and the infrastructure layer of your Internal Developer Platform (IDP).

A mature platform engineering workflow with Score looks like this:

The developer's workflow is: write score.yaml, test locally with score-compose, commit. Everything after the commit is the platform's responsibility. The developer never writes a Kubernetes manifest.

Integration with Platform Tooling

Score is designed to integrate with the broader platform ecosystem:

• Backstage: Score files can be registered as catalog entities. Backstage scaffolder templates can generate score.yaml stubs for new services, giving developers a correct starting point.
• GitOps (ArgoCD / Flux): The generated Kubernetes manifests are committed to a GitOps repository. ArgoCD or Flux applies them to the cluster. Score generates the input to the GitOps pipeline.
• Humanitec Platform Orchestrator: Humanitec (the company behind Score) builds a commercial platform orchestrator that natively consumes score.yaml files and handles resource resolution dynamically across environments.

Limitations to Know

Score is not a silver bullet. Understanding its limitations is as important as understanding its strengths:

• Resource type coverage: Score's built-in resource types (postgres, redis, dns, route) cover common cases. Unusual or proprietary infrastructure requires custom resource type definitions, which adds implementation complexity.
• Debugging generated output: When something goes wrong in production, engineers must trace the issue back through the generated manifest to the Score file. Adding this indirection layer means teams need tooling and discipline to keep the generated artifacts readable and auditable.
• Adoption friction: Score only delivers value when both developers and platform teams adopt it. A hybrid environment — some services using Score, others using raw Helm charts — can be harder to manage than a fully committed approach in either direction.