Validate schemas without relying on source filenames

internal/validate/schema_portability_test.go

@@ -0,0 +1,46 @@
+package validate
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"testing/fstest"
+
+	"go.maximumdirect.net/promptkit/internal/domain"
+)
+
+func TestSchemaPreparationDoesNotDependOnSourceName(t *testing.T) {
+	for _, test := range []struct {
+		name, schema string
+		rejected     bool
+	}{
+		{"source name", `{"$defs":{"v":{"type":"integer"}},"$ref":"schema.json#/$defs/v"}`, true},
+		{"unused source name", `{"$defs":{"v":{"$ref":"schema.json"}}}`, true},
+		{"fragment", `{"$defs":{"v":{"type":"integer"}},"$ref":"#/$defs/v"}`, false},
+		{"explicit ID", `{"$id":"https://example.invalid/schema.json","$defs":{"v":{"type":"integer"}},"$ref":"schema.json#/$defs/v"}`, false},
+		{"embedded ID", `{"$defs":{"v":{"$id":"child","type":"integer"}},"$ref":"child"}`, false},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			dir := t.TempDir()
+			if err := os.WriteFile(filepath.Join(dir, "schema.json"), []byte(test.schema), 0600); err != nil {
+				t.Fatal(err)
+			}
+			fsys := fstest.MapFS{"schema.json": &fstest.MapFile{Data: []byte(test.schema)}}
+			for i, validator := range []ValidationPreparer{NewStandardValidator(dir), NewFSValidator(fsys, ".")} {
+				t.Run(fmt.Sprint(i), func(t *testing.T) {
+					_, err := validator.PrepareValidation(context.Background(), domain.OutputContract{Format: domain.FormatJSON, ValidationMode: domain.ValidationJSONSchema, SchemaPath: "schema.json"})
+					if test.rejected {
+						if err == nil || !strings.Contains(err.Error(), "source location") {
+							t.Fatalf("expected source-independent rejection, got %v", err)
+						}
+					} else if err != nil {
+						t.Fatal(err)
+					}
+				})
+			}
+		})
+	}
+}

internal/validate/standard_validator.go

@@ -424,6 +424,20 @@ func (l externalSchemaLoader) Load(resourceURL string) (any, error) {
 }
 
 func compileSchemaDocument(ctx context.Context, resourceURL string, document any) (*jsonschema.Schema, error) {
+	root, err := compileSchemaAt(ctx, resourceURL, document)
+	if err != nil {
+		return nil, err
+	}
+	// File and private-FS retrieval URLs are never available to the provider.
+	// Compiling again under an unrelated HTTPS base detects reliance on those
+	// aliases without rewriting references or duplicating URI resolution rules.
+	if _, err := compileSchemaAt(ctx, "https://promptkit.invalid/submitted-schema", document); err != nil {
+		return nil, fmt.Errorf("schema is not independent of its source location; use fragment references or explicit schema IDs: %w", err)
+	}
+	return root, nil
+}
+
+func compileSchemaAt(ctx context.Context, resourceURL string, document any) (*jsonschema.Schema, error) {
 	if err := ctx.Err(); err != nil {
 		return nil, err
 	}