Final Answer Checklists

Use these when the user asks "which fields" or the issue clearly matches a product surface. They are the short-answer anchors too: do not rely on the model to infer these exact names from SKILL.md rules alone.

• Database Monitoring: say Database Monitoring requires all five target span attributes to be present: db.system, db.name, db.operation, db.statement, and net.peer.name. Treat db.system.name, db.namespace, db.operation.name, db.query.text, server.address, and network.peer.address as source/newer names that may need mirroring, not replacements for the target set.
• Span Metrics DB bridge: use top-level spanMetrics.transformStatements before spanmetrics, not only spanMetrics.dbMetrics.transformStatements. Include the symptom strings db_calls_total, calls_total, and db_namespace: DB-only placement can make db_calls_total get db_namespace while normal calls_total stays blank. The fallback chain is db.name, server.address, network.peer.name, net.peer.name, network.peer.address, then db.system.
• Fleet transform rendering: name both Fleet/UI span_metrics.transform_statements and rendered Helm spanMetrics.transformStatements; verify the rendered ConfigMap, not only values or UI JSON.
• Span Metrics labels: span_name, service_name, span_kind, status_code, collector_instance_id / collector.instance.id, http_method, host_name, k8s_cluster_name, cgx.transaction, cgx.transaction.root, db_namespace, db_operation_name, db_collection_name, db_system, application_name, and cx_subsystem_name.
• Span Metrics cardinality: avoid url.full, k8s.pod.ip, and blanket k8s.pod.name dimensions by default. Use templated http.route for HTTP grouping; add k8s.pod.name only for a concrete writer-collision case.
• Span Metrics writer identity: always include collector.instance.id / collector_instance_id or another writer-unique dimension. For collector-contrib v0.136.0 through v0.151.x, name --feature-gates=+connector.spanmetrics.includeCollectorInstanceID; for v0.152.0+, verify collector.instance.id is emitted.
• Span Metrics status-code bridge: always name spanmetrics.statusCodeConvention.useOtelPrefix. This changes generated RED metric data points from status.code to otel.status_code, so bridge in the metrics pipeline after spanmetrics, not trace spanMetrics.transformStatements or traces before spanmetrics. Raw values are status.code=Error/Ok/Unset and otel.status_code=ERROR/OK/missing, including otel.status_code=OK; do not match Prometheus STATUS_CODE_* labels in collector-side bridges.
• Span Metrics overflow: otel.metric.overflow="true" is the raw OTLP data-point attribute and otel_metric_overflow="true" is the Prometheus-normalized label for the same expected fallback series, not a bug. Reduce source cardinality first, for example normalize UUIDs out of span names and prefer http.route over raw url.path; do not present raising the cap as the primary fix.
• HTTP metric method: stable/current/new OpenTelemetry HTTP server metric data points should use http.request.method; Coralogix Span Metrics may still need a bridge for legacy http.method because connector dimensions read that source internally.
• Transactions timing: cgx.transaction / cgx.transaction.root are derived by CoralogixTransactionSampler; do not set them manually as the fix. Route-aware span names / http.route must exist before that sampler; collector rewrites after derivation can be too late.
• Infra Explorer pod identity: resource-scope k8s.cluster.name, k8s.namespace.name, k8s.pod.name, k8s.node.name, k8s.container.name, service.name, and one owner such as k8s.deployment.name, k8s.statefulset.name, k8s.daemonset.name, k8s.job.name, k8s.cronjob.name, or k8s.replicaset.name.
• EC2 Name tag: the target identity is resource-scope host.name; check EC2 tag/resource detection in logs and Resource Catalog plus ec2:DescribeTags.
• Lambda metadata: cx_metadata.* is product metadata. Name CX_LOGS_METADATA_ENABLED, CX_LOGS_METADATA_INCLUDE_TRACE_REF, and CX_LOGS_RESOURCE_BUILT_IN_ATTRIBUTES; warn about serverless dashboards and trace/log correlation.
• Custom metric normalized labels: service_instance_id is the Prometheus / Target Allocator normalized spelling of service.instance.id, not a separate identity dimension to keep with different values. Choose one canonical service or writer identity and align or remove the duplicate source.
• Exporter routing scope: cx.application.name / cx.subsystem.name are Coralogix proprietary routing attributes, not OTel-standard semconv. application_name_attributes and subsystem_name_attributes read resource attributes only. If a value is at log-record scope, recommend setting it at resource scope with SDK / resource / resourcedetection, or copying it up with OTTL; exact syntax belongs to opentelemetry-ottl, while core owns routing configuration. For env-var answers name OTEL_RESOURCE_ATTRIBUTES. Say core when naming the routing owner.
• Java instrumentation handoff: if the user asks whether semconv owns full Java auto-instrumentation setup, answer no, and name OTEL_EXPORTER_OTLP_ENDPOINT, OTEL_RESOURCE_ATTRIBUTES, and opentelemetry-instrumentation as the setup owner.
• Error Tracking timeout gaps: missing http.response.status_code can be normal when no HTTP response exists. exception.type / span error status are fallback signals; mapping them to a synthetic HTTP status is a Coralogix/product workaround, not standard OTel semconv.
• Resource Catalog inventory: normal k8sattributes / resource / transform adds correlation attributes but does not create inventory. Self-managed inventory needs k8sobjects, resourcedetection/resource_catalog, coralogix/resource_catalog, and x-coralogix-ingress: metadata-as-otlp-logs/v1. For freshness or stale metadata, mention resource_ttl, interval, otel.entity.interval, and schema_url.
• Pure OTTL handoff: if the user asks for an exact set(...) statement, where clause, or OTTL context, name opentelemetry-ottl as owner; semconv only names source and destination scopes.
• Other short answers: Azure cloud.platform is a value rename (azure_vm / azure.vm, azure_aks / azure.aks); timeout status mapping from exception.type is a Coralogix workaround; multi-gateway tail sampling / spanmetrics needs traceID load balancing.