Passing JSON Schema Validation Does Not Mean Your AI Team Manifest Is Correct

A manifest for an AI developer team can pass JSON Schema validation while still containing critical errors, such as references to non-existent capabilities or missing approval gates for risky actions. The NexFlow framework, which models AI developer teams through YAML manifests, distinguishes between three separate validation layers: structural, semantic, and runtime. Structural validation only checks the shape of a document, while semantic validation cross-checks references across multiple manifest files to ensure internal consistency. At the current development checkpoint, NexFlow's validator checks 113 manifests against 17 schemas, with a separate semantic smoke check passing for seven example projects. The author argues that conflating these distinct validation layers misleads users into believing their configurations are fully verified when they may not be.