Completed work
synthkit
372 completed tasks.
SKT-0228Derive only the deterministic correlation SpanID when the app discards the other fieldsSKT-0226Reject unsupported database trace identity engines instead of defaulting to PostgreSQLSKT-0225Restore hosted CI renderer identity proof across supported architecturesSKT-0210Reject the legacy throttle switch on causally bound agent runtimesSKT-0209Drive database observability from the same causal state as a bound databaseSKT-0209.01Bind PostgreSQL database observability to declared causal resource stateSKT-0209.02Bind MySQL database observability to declared causal resource stateSKT-0220Reconcile completed causal task checkboxes against retained delivery evidenceSKT-0168.09.32Sort map-order iterations that make base output nondeterministicSKT-0168.09.33Derive a bound resource's demand from upstream admitted callsSKT-0168.09.34Name a worker entry's consumer root span for its queue operationSKT-0168.09.35Keep unaffected span-metric rows on the legacy path in faulted windowsSKT-0168.09.36Stop parent hops inflating when a held pool's dependency is also calledSKT-0168.09.37Keep healthy DB child spans inside their parent spanSKT-0168.09.14Keep request outcome consistent when an error_spike fails a downstream hopSKT-0168.09.18Configure an app workload's success status and error status and code mixSKT-0168.09.28Keep a background error rate on causally bound Lambda functionsSKT-0168.09.29Mint queue-driven app requests from the resource's whole-event carrySKT-0168.09.31Apply gRPC span status rules to causal gRPC hopsSKT-0189Assert causal resource modes in the live scenario checkSKT-0195Count caller resends of a callee's subtree failures as demandSKT-0197Scale a client span-metric row only by its own caller's faulted hopsSKT-0199Stop a rule-only parent inflating when it calls several slowed children in sequenceSKT-0200Keep untouched services' trace spans at healthy duration on faulted requestsSKT-0204Give background and causal failures their error fields without empty values on healthy linesSKT-0185Bind DynamoDB tables to causal quota resources per operation and capacity unitSKT-0201Scope live lambda and SQS resource checks to account and regionSKT-0168.09.15Derive request latency and errors from resource state and declared faults across every signalSKT-0168.09.16Ramp an incident to its level and drain it after the incident endsSKT-0194Attribute TestEngineExactReadClientDeadlineCause failures under raceSKT-0196Keep a joined worker's trigger when one tick joins more than the trigger book holdsSKT-0198Scope lambda and SQS causal consumer identities to account and regionSKT-0168.09.30Propagate trace context from an app hop into the worker it triggersSKT-0190Align gateway export log status with causal request failuresSKT-0192Expose each causal resource's bound consumers in the control schemaSKT-0191Apply the resource-kind check to live control-plane incidentsSKT-0193Expand axis wildcards in live runtime incidentsSKT-0168.09.17Draw per-request values shared by every line of a log bundleSKT-0182.03Teach the causal form in blueprint authoring docs, the skill and an example fault catalogueSKT-0182.04Warn when a blueprint scenario scales rates without a declared causeSKT-0184Bind rds, docdb and agentcore constructs to causal resourcesSKT-0186Classify agento11y-provision TLS whole-closure test failure under loaded race gatesSKT-0187Render time.Duration blueprint fields as durations in the generated schemaSKT-0188Reject a resource-axis fault aimed at a resource of the wrong kindSKT-0168.09.19Model periodic batch arrivals and scheduled drainsSKT-0183Tighten unslashed archtest allowlist prefixesSKT-0182.01Sweep shipped blueprints for failure scenarios and plan the causal migrationSKT-0168.09.23Let histogram observation counts follow the load curve with a label subsetSKT-0168.09.24Give each broker queue its own curveSKT-0168.09.26Let a scheduled Lambda's duration follow the load curve at constant invocationsSKT-0168.09.27Freeze a heartbeat gauge while a declared incident stalls its collectorSKT-0168.02.07Capture and emit Lambda and DynamoDB throttle metricsSKT-0168.09.20Give a single series its own time-zone business profileSKT-0168.09.25Emit DynamoDB UpdateItem operationsSKT-0168.09.21Emit the current time as a gauge valueSKT-0168.09.22Scale scheduled Lambda duration with batch sizeSKT-0168.02.08Capture scheduler metrics and model scheduled-function cadenceSKT-0168.09.01Scale app custom-metric rates and values per enum label valueSKT-0168.09.02Derive app gauges as the stock of a counterSKT-0168.09.03Give app workloads their own diurnal profileSKT-0168.09.04Drive Lambda load and duration spread from the business curveSKT-0168.09.05Drive SQS arrivals and receives from the business curveSKT-0168.09.06Drive DynamoDB operations from the business curveSKT-0168.09.07Make the app workload request error rate configurableSKT-0168.09.08Allow a per-service load profile overrideSKT-0168.09.09Attribute a downstream service's error rate to its own hopSKT-0168.09.10Let one declared incident also drive Lambda and Solace constructsSKT-0168.09.11Resolve an incident target on the axis that declares its modeSKT-0168.09.12Commit tests for shared-name incident resolutionSKT-0168.09.13Let a shape value's incident level ignore the load curveSKT-0179Represent opt-in CloudWatch bases in the machine-readable signals catalogueSKT-0180Do not fail readiness on quiet request-driven delivery lanesSKT-0181Classify two load-sensitive gate failures in provisioning testsSKT-0168.06Declare sourced AWS services in the authorized private demoSKT-0168.02.06Add observed Lambda Logs delivery period observationsSKT-0168.02.04Implement the bounded provisioned Redshift CloudWatch snapshot subsetSKT-0168.02.05Implement the bounded CloudWatch Logs intake metric subsetSKT-0177Reconcile containerd runtime identity documentation with the closed descriptor contractSKT-0178Protect the verified main-SHA image selected by downstream deploymentsSKT-0176Separate chart startup health from first-delivery readinessSKT-0168.02.01Implement captured Lambda CloudWatch function metricsSKT-0168.02.02Implement captured SQS CloudWatch queue metricsSKT-0168.02.03Implement captured DynamoDB CloudWatch table metricsSKT-0171Preserve explicit zero intensity in scheduled blueprint incidentsSKT-0175Reject non-finite incident intensity valuesSKT-0173Retain HTTP-client interruption causes inside engineReadExactSKT-0174Bring custody forecast race tests below the unchanged hosted package timeoutSKT-0153.03.04.04Keep the underlying error when the engine reports an exact-read conflictSKT-0172Preserve fractional Solace publish and drain capacity across observation windowsSKT-0168.05Source and emit MWAA DAG and task outcome metricsSKT-0153.03.04.02Refuse an Agent Observability apply whose teardown the custody budget would refuseSKT-0168.04Verify solace_prometheus_exporter metric names and add a Solace broker constructSKT-0170Allow a zero-baseline custom metric to rise under a scoped incidentSKT-0166Investigate TestEnvSurfaceAligned failing in gate sweeps of a fresh cloneSKT-0169Make the blueprint schema generator order failure modes deterministicallySKT-0153.03.04.03Tighten engine lifecycle tests that accept any engine errorSKT-0168.01Add the demo's customer-identifying terms to the local forbidden-words fileSKT-0153.03.05Add source-built manual Agent Observability provisioning applicationSKT-0153.03.04Execute the typed create-only Agent Observability lifecycle against bound custodySKT-0162Run timing-sensitive HA subprocess tests in an isolated race invocationSKT-0164Bind the non-HA control listener before logging so tests need no reserved portSKT-0167Stop the k8s_profiling construct emitting full-load profiles at volume 0SKT-0163Make ai_agent honour volume_multiplier 0 as a pauseSKT-0147Make an explicit volume_multiplier of 0 pause emission instead of running at full volumeSKT-0077Emit the non-network Datadog standalone-host breadth from the committed verdictSKT-0160Let the app pipeline emit metadata-only thinking parts through the explicit part kindSKT-0161Replace reserve-and-close test ports in cmd/synthkit HA and Fleet fixtures with child-bound listenersSKT-0157Preserve native generation part kind when captured content is intentionally emptySKT-0159Keep chart CI fixture contract checks compatible with routine reusable pin updatesSKT-0110Diagnose GHCR digest-not-found in ARM64 image publicationSKT-0136Bound one-shot validation and preflight by an absolute deadlineSKT-0153.03.04.01Add revision-gated bounded state capacity for complete typed checkpointsSKT-0153.05.03.01Collect pure full-set agent declarations with native child identity paritySKT-0153.03.03Add local provisioner custody and structured approval foundationsSKT-0153.05.02Implement the bounded app agent-pipeline executorSKT-0153.03.02Add rule collection action and suite management transportSKT-0153.02Refresh the Agent Observability contract to SDK 75a1c548 and backend bc963fb09SKT-0153.03.01Add bounded evaluator and hook management transportSKT-0153.05.01Freeze shared agent declaration and conversation rating ABISKT-0153.07Add a fictional regulated-document authoring corpusSKT-0153.08Source and model Bedrock application inference profiles per agentSKT-0154Add a generic app custom-metrics mechanism for blueprint-declared business metricsSKT-0153.01Capture customer-stack Agent Observability baselines before the contract refreshSKT-0155Stop the local forbidden-words guard false-positive on an SM identifierSKT-0151Admit live Synthetic Monitoring per blueprint instead of re-validating the whole snapshotSKT-0152Bring cmd/synthkit-control-dash race runtime back well under the 600 s test timeoutSKT-0138Publish a bounded operational manifest of intended delivery lanes and cadencesSKT-0139Teardown every persisted owned Fleet collector independently of current selectionSKT-0148Bring the HA composition root to parity for live-restored blueprintsSKT-0149Let Synthetic Monitoring cleanup run after SM is disabled or deselected, and package it for KubernetesSKT-0134Keep console SVG icons legible in dark themeSKT-0137Expose an unambiguous leadership-session identity on operational telemetrySKT-0140Add ownership-bound preview and explicit cleanup for provisioned Synthetic Monitoring resourcesSKT-0141Support deterministic dashboard UID namespaces and matching artifact inventorySKT-0146Admit Synthetic Monitoring constructs in live blueprint apply through the boot SM handoffSKT-0150Fix the dashboards example push loop and the synthkit-dash flag docs driftSKT-0142Apply and remove one namespaced blueprint live on a running processSKT-0143Disable construct kinds from the first tick with a startup settingSKT-0144Expose per-blueprint emitting flags on the public readiness probeSKT-0145Document an emission-only local run against a Grafana Cloud stackSKT-0087.04Add an NVIDIA DCGM exporter construct for GPU telemetrySKT-0135Bridge standard operational logs across HA role transitionsSKT-0079Add a selectable OTel-receiver pod-log resource shape and declare otel-receivers modelledSKT-0101Profile and reduce the per-process memory baselineSKT-0103Refresh the control console design and extract shared design tokensSKT-0132Reconcile legacy HTTP Synthetic Monitoring histogram buckets with the current agentSKT-0133Align onboarding, contributor workflows and documentation with current runtimeSKT-0087.25Add a shared GPU clock and thermal/power constraint modelSKT-0089Model Synthetic Monitoring network check types (ping, DNS, TCP, traceroute)SKT-0130Let the chart run Synthetic Monitoring emission with HA and named state togetherSKT-0087.26Tighten the HGX domain-node parent assertionSKT-0088Give each Synthetic Monitoring private probe its own locationSKT-0091Make nettopo_spoke_down take down a subset of spokes, not all of themSKT-0126Fence snapshot publication on acquired provisioner storesSKT-0090Cut a release so the network demo host stops bind-mounting blueprintsSKT-0131Wire named Fleet state into the Kubernetes HA chartSKT-0099.01Wire diskless SM provisioning into the existing named-state chartSKT-0129Project declarative git blueprint sources without overwriting instance-owned stateSKT-0099Make Synthetic Monitoring provisioning work without a shared volumeSKT-0127Synchronize the SM lost-response fixture observationsSKT-0125Keep operator-owned-state chart proof compatible with current HelmSKT-0124Investigate missing Sigil aggregate observation on between-stage cancellationSKT-0097Add an HA mode to the Helm chartSKT-0100Add Fleet Management cleanup for handoffs and teardownSKT-0112Keep the control-dash race package inside the test timeoutSKT-0116Bound release-please closed-connection failure on mainSKT-0122Resolve generic-api-key false positive on ordinary teardown documentationSKT-0123Classify intermittent Sigil three-stage deadline proof failureSKT-0105Let a deployment declare managed console features that the UI locks with a reasonSKT-0120Build current embedded UI before Go validationSKT-0121Synchronize HA loss proof with actual HTTP handler observationSKT-0094Add a kubernetes state backend for control stateSKT-0098Serve the control console under a runtime base pathSKT-0114Diagnose the flaky network probe bound preflight testSKT-0093Make readiness standby-aware in HA modeSKT-0119Ship synthkit-dash in the runtime image so a deployment can generate dashboards without a source checkoutSKT-0118Clear x/net HIGH vulnerability from capture image publicationSKT-0006.03k3d capture lab: real k8s-monitoring egress into a capture receiver, local one-shot plus nightly CISKT-0019Migrate the repo task surface to just and retire Makefiles and ad-hoc scriptsSKT-0049Restore otlp-native to the lab deployment once a fixed image shipsSKT-0109Resolve late Faro POST in composed lease HA gateSKT-0092Add a lease-based HA mode with crash-only fencingSKT-0113Diagnose the flaky HA planned-release lifecycle testSKT-0115Resolve skcapture Trivy publication failures on both platformsSKT-0108Align local environment surface after lease HA landingSKT-0111Archive local Go reference sources outside module traversalSKT-0107Keep live app trace identities unique across restarts without destabilizing inventorySKT-0095Anchor construct phase scheduling to the wall clockSKT-0096Restrict git source token variables to an allowlisted prefixSKT-0104Add a one-shot validate mode for blueprint setsSKT-0106Make app agent-flow inventory deterministic for full dump paritySKT-0087.02Add a shared syslog-ingest log mechanic matching Alloy's syslog receiversSKT-0087.16Model a clustered site gateway collector pool and telemetry-gap signalsSKT-0087.23Add an HGX 8-GPU shape to the GPU compute topology fixtureSKT-0087.01Add a GPU compute topology fixture shared by the AI Factory constructsSKT-0087.17Add AI Factory platform-automation workflow tracesSKT-0086Prune change-detector Go testsSKT-0081Prune change-detector unit testsSKT-0085Make the app default-wire parity golden platform-independentSKT-0084Model per-node telemetry adoption in the app workloadSKT-0083Add a control-plane instructor layout to synthkit-control-dashSKT-0082Port server-side Infinity actions into synthkit-control-dashSKT-0074Repair stale lab acceptance predicates after producer projectionSKT-0078Ground or reject the 75 system.net standalone-host candidatesSKT-0080Convert the producer-coverage ratchet into a claimed ledgerSKT-0075Cover the final minute in runner incident test windowsSKT-0076Model OTLP pod logs and right-size producer coverage after live receiver proofSKT-0070Close standalone Docker machine identity gaps from the independent cAdvisor captureSKT-0072Replace frozen Datadog receiver system values with grounded deterministic mechanicsSKT-0073Classify every standalone Datadog host capture family before further implementationSKT-0055Model the Datadog agent receiver path from independent post-gateway capturesSKT-0068Close the six proven P3 producer and label envelope omissionsSKT-0060Read dump metric names against documented signal formats with explicit unresolved evidenceSKT-0069Keep secret scanning active for captured Datadog attribute namesSKT-0056Emit the captured OTel Collector with Prometheus exporters permutationSKT-0058k3d image import reports success while the nodes loaded nothingSKT-0059Scope signal identity by reviewed producer, and fail loud when no producer is comparableSKT-0067Make the parallel-isolation test wait for observed dropped ticksSKT-0054Resolve the control-plane PENDING rows from observed scheduler, controller-manager and etcd evidenceSKT-0066Validate P3 job identity after corpus producer projectionSKT-0061Pass explicit Kubernetes contexts throughout the local capture matrixSKT-0064Label lab receiver receipt counts by their actual unitsSKT-0020.04Rancher capture environment and its first corpus captureSKT-0053Merge the second AWS, Azure and GCP captures into the corpus by cumulative unionSKT-0062Preserve synthetic metric shapes per producer during corpus comparisonSKT-0063Correct captured Kubernetes control-plane histograms and scrape envelopesSKT-0065Validate cumulative corpus coverage against reviewed routes instead of frozen snapshot countsSKT-0057Expose bounded Fleet collector configuration receipt evidenceSKT-0012.11Publish the skcapture image on release tags and verify it alongside the main imageSKT-0051Emit the captured agento11y_eval enqueue and judge-cost familiesSKT-0052Two fixed-tick dump runs differ in logs and spans: find the unseeded drawSKT-0012Validate skcapture and skforge against the lab Kubernetes environmentsSKT-0002sigil: re-capture a live heuristic evaluator to confirm its series shapeSKT-0020Multi-cloud capture environments: model synthkit against every major managed KubernetesSKT-0020.02AKS capture environment and its first corpus captureSKT-0020.03GKE capture environment and its first corpus captureSKT-0020.06Port the EKS capture environment into synthkit-terraformSKT-0007.09Capture and decide csp_azure and csp_gcp OTel-native formSKT-0007OTLP-native emission parity across the catalogSKT-0020.01Stand up the synthkit-terraform repository and freeze its module shapeSKT-0012.05skcapture and skforge: cover the in-cluster RBAC path and a non-EKS substrateSKT-0010Make the signal-fidelity comparator trustworthy and give the gate teethSKT-0010.21Bring the signal-fidelity report under its documented 27,212-line boundSKT-0012.10skforge: fail clearly on a non-EKS or undetermined capture instead of emitting an AWS placeholder skeletonSKT-0014.03Deploy synthkit to the lab cluster as an ArgoCD applicationSKT-0020.05Make per-substrate divergence first-class in the comparatorSKT-0010.02Record real instrument types in both corpus producersSKT-0011User-acceptance round: agent-run scenarios on a clean machine and a fresh Grafana Cloud stackSKT-0010.08Fold classic-histogram component suffixes in the gcx read-back producerSKT-0004sigil: -once -dump generation/score COUNTS vary run-to-run — decide whether that is by design or a determinism defectSKT-0010.07Correct two emitter shape defects the coverage audit surfacedSKT-0021.01118 instrument_mismatch gaps are all reality=unknown: decide capture enrichment or absent-evidenceSKT-0006Signal fidelity audit: validate emitted telemetry shapes against real captured telemetrySKT-0007.08Emit the CloudWatch group OTel-native under the metric-stream naming lawSKT-0013Model the documented k8s-monitoring deployment permutationsSKT-0007.07Emit beyla_agent and envoy_gateway OTel-nativeSKT-0010.20Correct kubelet runtime operation labels from upstream sourceSKT-0010.14Move the log-source keying across the receiver, the dumper and ParseDump togetherSKT-0012.06skcapture: detect the platform products its well-known tables missSKT-0012.08skcapture: detect the provider on a Karpenter-only EKS clusterSKT-0008.02Emit target_info for the app workloadSKT-0012.09Decide the RBAC grant skcapture needs to read cluster identity in-clusterSKT-0014.02Correct .env.example and the container detection it over-promisesSKT-0012.07skcapture: the secret-data flags are inert, but the docs and RBAC promise themSKT-0013.02Model the Prometheus operator remote-write envelopeSKT-0042Recapture mixed-substrate corpus input with unambiguous provenanceSKT-0013.03Refresh the k3d lab to the k8s-monitoring version the estate actually runsSKT-0050Four of the lab's seven selected blueprints emit no synthetic telemetrySKT-0031Make all shipped blueprints individually deployable and verifiable by runtime identitySKT-0043Sigil is now Agent Observability: re-verify endpoints, auth and payload formatSKT-0014.01Validate the chart end to end: deploy it in k3d and check what it emitsSKT-0010.06Emit the CloudWatch families the coverage audit verdicted should-emitSKT-0008.01Derive span-metric status_code from the trace lane instead of hard-coding itSKT-0008.03Drive the span-metric error fraction from ledger outcomesSKT-0010.19Carry direct producer identity through privacy-safe corpus comparisonSKT-0021Close the reality coverage gaps, by cause rather than by countSKT-0021.02375 CloudWatch families reality publishes and synthkit does not emitSKT-0021.0343 k8s plus 2 k8s-addon gaps, and 19 host findingsSKT-0044Blueprint option: emit only what the k8s-monitoring default allow-lists keepSKT-0045Teach the fidelity comparator what an unknown metric type meansSKT-0046Model the OTLP-native label set, not just the Alloy-converted oneSKT-0047Restyle the control UI onto design system v2SKT-0048Model observed span resource attributes: cloud.* is a minority, and host.name is the collectorSKT-0033Make control UI, inventory, diagnostics, and scaling self-verifyingSKT-0034Make scenario activation effects observable in emitted dataSKT-0035Make upgrade, rollback, and snapshot persistence reproducible from public releasesSKT-0039Verify the optional signal lanes with the credentials now availableSKT-0040Make k8s corpus histogram bucket evidence consistentSKT-0041Convert storage operation duration only from captured histogram boundsSKT-0010.13Model pods with no Deployment owner and no nodeSKT-0013.10Make both pod-log transports explicitly declared and guardedSKT-0015Emit high-DPM and high-churn series on demand, for testing detection toolingSKT-0030Make the clean-container deployment path executable from shipped instructionsSKT-0032Fix valid custom blueprint upload returning HTTP 400SKT-0036Make generated dashboards self-contained and panel-verifiable on a fresh stackSKT-0037Turn troubleshooting documentation into an executable symptom matrixSKT-0038Make marketplace installation instructions executable from a clean hostSKT-0027The only-in-synth value rule is too blunt: deployment-specific values are absent evidenceSKT-0028Three exemption rules exempt a likely defect or modelling gap, not capture narrownessSKT-0029Exemption drift aborts before the report, hiding every co-occurring contradictionSKT-0010.05Fail CI on fidelity contradictions, report coverage gapsSKT-0014.04Support Gateway API HTTPRoute in the chart, and reconcile exposure with itSKT-0022k8s-addons stamp an invented node label and omit sourceSKT-0023k8s pollers stamp source on kube-proxy families where reality has noneSKT-0024CloudWatch stamps TargetGroup on load-balancer metrics and tag_VpcId on info seriesSKT-0025Reclassify subset label values as coverage gaps, not contradictionsSKT-0026A two-direction label divergence is reported as both a contradiction and a gapSKT-0010.18Scope label comparison by producer, so a multi-producer union is not a contradictionSKT-0010.17go_gc_duration_seconds reports as a gauge against captured summary evidenceSKT-0015.06high_dpm is inert: correct the floor rule so a default-cadence construct is sped upSKT-0017Readiness counts lanes with credentials, not lanes a blueprint feedsSKT-0018OTLP-logs delivery is invisible to push status because its observer is not attachedSKT-0007.05Emit app and ai_agent OTel-nativeSKT-0007.06Emit host OTel-native via the hostmetrics shapeSKT-0010.15Pair a source-declaring log lane in the comparator instead of reporting it uncoveredSKT-0010.16Refresh the k3d metric corpus from the corrected capture, with a lab-artefact filterSKT-0013.09Correct and exercise the k8s_cluster classic Loki pod-log laneSKT-0015.04Make high_dpm a floor, not a forced cadenceSKT-0015.05Add declarable pod lifecycle churn, coherent across every signalSKT-0016CI e2e fails: 131 CloudWatch families declared by -dump are never receivedSKT-0007.04Emit k8s_cluster OTel-native: the base blueprint surfaceSKT-0013.05Decide which Loki pod-log shape synthkit models, before the permutation is capturedSKT-0013.08Correct the Loki pod-log emitter to the captured stream shapeSKT-0015.01Make the DPM floor a per-blueprint opt-in override with a configurable ceilingSKT-0015.02Make series churn a declarable rate on the constructs that already model identity lifecycleSKT-0015.03Ship a high-DPM reference blueprint and make its cost visible before it runsSKT-0013.06Stop the capture receiver recording inferred evidence as observedSKT-0013.07Re-examine every prior capture now that Loki decode worksSKT-0010.10Separate the reading tool version from corpus merge identitySKT-0013.01Capture the OTel-native receiver permutation and resolve SK-85SKT-0013.04Run the k3d lab as a parallel permutation matrix with one combined reportSKT-0010.12Elide capture-instance identity values in the k3d lab producerSKT-0014Ship a Helm chart for synthkitSKT-0010.09Classify log entries by shape, not by an absent source labelSKT-0010.11Emit le in the Prometheus v3 float-normalised form across the catalogueSKT-0007.03Guard the scrape-only OTel verdicts with an architecture testSKT-0012.01skcapture: take cluster identity from the cluster, not the operator kubeconfigSKT-0012.02skcapture: read real node pool identity instead of synthesising groups per instance typeSKT-0012.03skcapture: stop the zero-secret default leaking through annotationsSKT-0012.04skforge: fix the mapper and coverage report misreporting what synthkit can modelSKT-0008Validate self-emitted span metrics against both real producers, and generate native histogram panelsSKT-0006.05Emit Kubernetes pod logs OTel-native (OTLP logs) alongside the Loki-native pathSKT-0010.01Comparator evidence semantics: absent evidence is a gap, never a contradictionSKT-0010.04Triage every coverage-gap metric into a recorded verdictSKT-0007.01Establish which catalog kinds have a real OTel-native metric formSKT-0007.02Close the OTLP metrics lane instrument-shape gapsSKT-0010.03Correct the real emission divergences the fidelity audit foundSKT-0006.04gcx live read-back: merge real EKS label names and values into the reality corpusSKT-0006.02Reality corpus format and report-only PR fidelity gateSKT-0006.01Canonical machine-readable telemetry inventory schema and JSON exportSKT-0009Upgrade Go toolchain to 1.27SKT-0005Operational readiness for first-time Grafana practitionersSKT-0005.14Make upgrades reproducible and rollback testableSKT-0005.11Expose delivery loss, queue pressure, and deployment sizingSKT-0005.19Start fresh deployments with no selected blueprintsSKT-0005.12Harden control-plane exposure and local file permissionsSKT-0005.09Make git blueprint sources and apply state explicitSKT-0005.10Model real deployment readiness in Compose and status APIsSKT-0005.17Validate mandatory live Grafana configuration before startupSKT-0005.06Complete and forward-test the shipped operational skillsSKT-0005.07Correct beginner documentation and control-plane examplesSKT-0005.08Make custom blueprint upload validation trustworthySKT-0005.04Provide one focused first-workload deployment pathSKT-0005.05Make marketplace installation truthful and relocatableSKT-0005.15Make the documentation site locally buildable or verifiableSKT-0005.16Make live blueprint and scenario mutations safe for API clientsSKT-0005.18Reduce always-loaded agent instructions to the portable contractSKT-0005.01Make clone-to-live defaults safe and explicitSKT-0005.02Repair canonical agent instructions and cross-harness skill syncSKT-0005.03Harden env helper scripts and secret-file replacementSKT-0001sigil: carry a turn's tool-results into the next turn's input, and per-subagent agent_name for claude-code childrenSKT-0003nodeexp: resolve the stale TODO(phase5) full-profile linux delta