Skip to content

Frontend Stack

The infrastructure.frontend_stack module defines CloudFront plus the S3 access-log bucket and the Glue database / Athena saved queries that make the access logs queryable.

API reference

infrastructure.frontend_stack

FrontendStack(scope, construct_id, *, api_url, api_id, waf_acl_arn, cf_waf_logs_location, regional_waf_logs_location, **kwargs)

Bases: Stack

CDK stack for the Serverless App frontend.

Provisions a private S3 bucket for static assets and a CloudFront distribution with OAC, HTTPS-only enforcement, and security response headers. WAF protection is provided by a WebACL ARN passed in from WafStack, which is always deployed in us-east-1.

This stack can be deployed to any region. When the target region differs from us-east-1, CDK bridges the WAF ARN cross-region automatically via SSM Parameter Store (enabled by cross_region_references=True in app.py).

Provision all frontend AWS resources.

Parameters:

Name Type Description Default
scope Construct

The CDK construct scope.

required
construct_id str

The unique identifier for this stack.

required
api_url str

The backend API Gateway URL, injected into config.json at deploy time.

required
api_id str

The backend API Gateway REST API ID, used to pin the CSP connect-src to the exact execute-api host instead of a region-wide wildcard.

required
waf_acl_arn str

ARN of the WAF WebACL from WafStack (always in us-east-1).

required
cf_waf_logs_location str

s3://…/ prefix of the CloudFront WebACL's WAF logs, for the Athena Glue table (computed by the Stage — see its docstring).

required
regional_waf_logs_location str

s3://…/ prefix of the regional (API Gateway) WebACL's WAF logs, for the Athena Glue table.

required
**kwargs Any

Additional keyword arguments passed to the parent Stack.

{}
Source code in infrastructure/frontend_stack.py
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
def __init__(
    self,
    scope: Construct,
    construct_id: str,
    *,
    api_url: str,
    api_id: str,
    waf_acl_arn: str,
    cf_waf_logs_location: str,
    regional_waf_logs_location: str,
    **kwargs: Any,
) -> None:
    """Provision all frontend AWS resources.

    Args:
        scope: The CDK construct scope.
        construct_id: The unique identifier for this stack.
        api_url: The backend API Gateway URL, injected into config.json at deploy time.
        api_id: The backend API Gateway REST API ID, used to pin the CSP connect-src
            to the exact execute-api host instead of a region-wide wildcard.
        waf_acl_arn: ARN of the WAF WebACL from WafStack (always in us-east-1).
        cf_waf_logs_location: ``s3://…/`` prefix of the CloudFront WebACL's WAF logs,
            for the Athena Glue table (computed by the Stage — see its docstring).
        regional_waf_logs_location: ``s3://…/`` prefix of the regional (API Gateway)
            WebACL's WAF logs, for the Athena Glue table.
        **kwargs: Additional keyword arguments passed to the parent Stack.
    """
    super().__init__(scope, construct_id, **kwargs)
    self._api_id = api_id
    self._cf_waf_logs_location = cf_waf_logs_location
    self._regional_waf_logs_location = regional_waf_logs_location

    apply_compliance_aspects(self)

    # ── KMS key ──────────────────────────────────────────────────────────
    # Used to encrypt the frontend S3 bucket and CloudWatch log group.
    # CloudWatch Logs requires the Logs service principal in the key policy.
    frontend_encryption_key = kms.Key(
        self,
        "FrontendEncryptionKey",
        description=f"KMS key for {self.stack_name} S3 bucket and log groups",
        enable_key_rotation=True,
        # See BackendApp.encryption_key for the rationale — automated
        # rotation, no dependent redeploys, 90-day compliance baseline.
        rotation_period=Duration.days(90),
        removal_policy=RemovalPolicy.DESTROY,
    )
    # Confused-deputy guard on the CMK's CloudWatch Logs service grant.
    # See ``grant_logs_service_to_key`` in ``nag_utils.py``.
    grant_logs_service_to_key(
        frontend_encryption_key,
        region=self.region,
        account=self.account,
        partition=self.partition,
    )

    # ── S3 access logging bucket ─────────────────────────────────────────
    # Receives S3 server access logs (asset bucket), CloudFront standard
    # access logs, and Athena query results. SSE-S3 (not SSE-KMS) because
    # neither S3 log delivery nor CloudFront standard logging support
    # KMS-encrypted target buckets. Built via the shared log-sink helper;
    # object_ownership=BUCKET_OWNER_PREFERRED keeps ACLs usable for CloudFront
    # standard logging (which delivers via ACL). 7-day lifecycle is tunable —
    # extend or tier to Glacier for longer retention (see "Audit stack and
    # log retention" in the README).
    access_log_bucket = create_sse_s3_log_bucket(
        self,
        "FrontendAccessLogBucket",
        suppression_reason=(
            "Access-log bucket — SSE-S3 (S3/CloudFront log delivery doesn't support "
            "KMS-CMK target buckets), self-logging would be circular, no versioning/"
            "replication for append-only transient logs"
        ),
        expiration_days=7,
        removal_policy=RemovalPolicy.DESTROY,
        auto_delete=True,
        object_ownership=s3.ObjectOwnership.BUCKET_OWNER_PREFERRED,
    )

    # ── S3 bucket ────────────────────────────────────────────────────────
    # Fully private — CloudFront OAC is the only allowed reader.
    # KMS-encrypted with server access logging to access_log_bucket.
    bucket = s3.Bucket(
        self,
        "FrontendBucket",
        block_public_access=s3.BlockPublicAccess.BLOCK_ALL,
        encryption=s3.BucketEncryption.KMS,
        encryption_key=frontend_encryption_key,
        enforce_ssl=True,
        server_access_logs_bucket=access_log_bucket,
        server_access_logs_prefix="s3-access-logs/",
        versioned=False,
        removal_policy=RemovalPolicy.DESTROY,
        auto_delete_objects=True,
    )

    # Expose the buckets for the audit stack to consume cross-stack — the
    # CloudTrail trail recording object-level data events on them lives in
    # AuditStack (a one-way dependency: audit -> frontend).
    self.bucket = bucket
    self.access_log_bucket = access_log_bucket

    # ── CloudFront response headers ──────────────────────────────────────
    # Custom ResponseHeadersPolicy (replaces the AWS-managed SECURITY_HEADERS)
    # adding HSTS + CSP on top of the four headers that policy provided.
    # See _build_response_headers_policy for the full rationale.
    response_headers_policy = self._build_response_headers_policy()

    # ── CloudFront distribution ──────────────────────────────────────────
    distribution = cloudfront.Distribution(
        self,
        "Distribution",
        default_behavior=cloudfront.BehaviorOptions(
            origin=origins.S3BucketOrigin.with_origin_access_control(bucket),
            viewer_protocol_policy=cloudfront.ViewerProtocolPolicy.REDIRECT_TO_HTTPS,
            cache_policy=cloudfront.CachePolicy.CACHING_OPTIMIZED,
            response_headers_policy=response_headers_policy,
        ),
        default_root_object="index.html",
        error_responses=[
            # Return index.html for 403/404 so SPA client-side routing works
            cloudfront.ErrorResponse(
                http_status=403,
                response_http_status=200,
                response_page_path="/index.html",
            ),
            cloudfront.ErrorResponse(
                http_status=404,
                response_http_status=200,
                response_page_path="/index.html",
            ),
        ],
        minimum_protocol_version=cloudfront.SecurityPolicyProtocol.TLS_V1_2_2021,
        web_acl_id=waf_acl_arn,
        enable_logging=True,
        log_bucket=access_log_bucket,
        log_file_prefix="cloudfront/",
    )

    # ── CloudWatch RUM + X-Ray ───────────────────────────────────────────
    # RUM collects browser telemetry (page loads, JS errors, fetch latency)
    # and — with enable_x_ray — emits a client-side trace segment that joins
    # the backend Lambda/API Gateway segments into a single X-Ray trace.
    # Guest (unauthenticated) browsers authenticate via Cognito Identity
    # Pool → STS AssumeRoleWithWebIdentity → scoped rum:PutRumEvents role.
    # The monitor ARN is constructed from the known monitor name so the
    # IAM role can reference it without a circular dependency on the
    # CfnAppMonitor resource.
    rum_identity_pool = cognito.CfnIdentityPool(
        self,
        "RumIdentityPool",
        allow_unauthenticated_identities=True,
        identity_pool_name=f"{self.stack_name}-rum",
    )
    # Pinned physical name (the IAM role below references the constructed
    # ARN): a future replacement-forcing property change collides with the
    # not-yet-deleted old monitor (CFN replacement is create-before-delete),
    # so such a change must also change the name in the same commit — see
    # the AppConfig profile note in backend_app.py.
    rum_monitor_name = f"{self.stack_name}-rum"
    rum_monitor_arn = f"arn:{self.partition}:rum:{self.region}:{self.account}:appmonitor/{rum_monitor_name}"
    rum_unauth_role = iam.Role(
        self,
        "RumUnauthenticatedRole",
        assumed_by=iam.FederatedPrincipal(
            "cognito-identity.amazonaws.com",
            conditions={
                "StringEquals": {"cognito-identity.amazonaws.com:aud": rum_identity_pool.ref},
                "ForAnyValue:StringLike": {"cognito-identity.amazonaws.com:amr": "unauthenticated"},
            },
            assume_role_action="sts:AssumeRoleWithWebIdentity",
        ),
        description=f"Guest role assumed by browser RUM clients for {rum_monitor_name}",
        inline_policies={
            "AllowPutRumEvents": iam.PolicyDocument(
                statements=[
                    iam.PolicyStatement(
                        actions=["rum:PutRumEvents"],
                        resources=[rum_monitor_arn],
                    )
                ]
            )
        },
    )
    cognito.CfnIdentityPoolRoleAttachment(
        self,
        "RumIdentityPoolRoleAttachment",
        identity_pool_id=rum_identity_pool.ref,
        roles={"unauthenticated": rum_unauth_role.role_arn},
    )
    rum_app_monitor = rum.CfnAppMonitor(
        self,
        "RumAppMonitor",
        name=rum_monitor_name,
        domain=distribution.distribution_domain_name,
        cw_log_enabled=True,
        # Enable custom events so the frontend can call cwr('recordEvent', ...)
        # for domain telemetry. Without this set to ENABLED, custom event
        # uploads are silently dropped at the data plane.
        custom_events=rum.CfnAppMonitor.CustomEventsProperty(status="ENABLED"),
        app_monitor_configuration=rum.CfnAppMonitor.AppMonitorConfigurationProperty(
            allow_cookies=True,
            enable_x_ray=True,
            # session_sample_rate is a CLIENT-SIDE knob: the RUM browser
            # client reads it to decide what fraction of sessions to record.
            # It does NOT bound the data plane — the public unauthenticated
            # identity pool above lets anyone mint guest credentials and call
            # rum:PutRumEvents directly, ignoring this rate entirely. So
            # lowering it is a legitimate-traffic COST lever only; the
            # adversarial-ingestion vector needs a Budgets / RUM-volume alarm,
            # not a smaller sample rate (see TODO.md "CloudWatch RUM").
            session_sample_rate=1.0,
            # CloudFormation's schema only accepts ["errors", "performance", "http"] here —
            # "interaction" is rejected as an invalid enum value despite being a real RUM
            # plugin. This server-side list is metadata used by the AWS-generated snippet,
            # not the live plugin loader. The actual plugin set is controlled by the
            # client-side `telemetries` array in frontend/index.html, which DOES include
            # "interaction" alongside the http tuple form. Keep these two lists divergent
            # on purpose; do not "sync" them.
            telemetries=["errors", "performance", "http"],
            identity_pool_id=rum_identity_pool.ref,
            guest_role_arn=rum_unauth_role.role_arn,
        ),
    )

    # CMK-encrypted log group for the BucketDeployment provider Lambda.
    # Passing log_group= here (instead of log_retention=) avoids the legacy
    # LogRetention singleton path and keeps every log group encrypted with
    # this stack's CMK.
    bucket_deployment_log_group = logs.LogGroup(
        self,
        "BucketDeploymentLogGroup",
        encryption_key=frontend_encryption_key,
        retention=logs.RetentionDays.ONE_WEEK,
        removal_policy=RemovalPolicy.DESTROY,
    )

    # Shared CMK-encrypted log group for all AwsCustomResource singletons in
    # this stack (RumMetricsDestination, RumExtendedMetrics, InvalidateCloudFrontCache).
    # CDK reuses one provider Lambda across every AwsCustomResource in a stack,
    # so a single log group serves all three.
    custom_resource_log_group = logs.LogGroup(
        self,
        "AwsCustomResourceLogGroup",
        encryption_key=frontend_encryption_key,
        retention=logs.RetentionDays.ONE_WEEK,
        removal_policy=RemovalPolicy.DESTROY,
    )

    rum_extended_metrics = self._wire_rum_metrics_extras(
        rum_app_monitor, rum_monitor_name, rum_monitor_arn, custom_resource_log_group
    )
    self._wire_rum_log_group_cleanup(rum_app_monitor, rum_monitor_name, custom_resource_log_group)

    # ── Deploy frontend assets ───────────────────────────────────────────
    # Uploads frontend/ to S3 and generates config.json with the API URL
    # and RUM client config injected at deploy time. Cache invalidation is
    # handled by a separate AwsCustomResource below — the BucketDeployment's
    # built-in `distribution=` parameter is intentionally not used because
    # its delete-time invalidation races with CloudFront's own deletion on
    # `cdk destroy` (aws/aws-cdk#15891).
    bucket_deployment = s3deploy.BucketDeployment(
        self,
        "DeployFrontend",
        sources=[
            s3deploy.Source.asset("frontend"),
            s3deploy.Source.json_data(
                "config.json",
                {
                    "apiUrl": api_url,
                    "rum": {
                        "appMonitorId": rum_app_monitor.attr_id,
                        "identityPoolId": rum_identity_pool.ref,
                        "region": self.region,
                        # Session attributes are attached to every RUM event
                        # in the session. Sourcing them from deploy-time
                        # config (rather than hardcoding in the HTML) lets
                        # multiple deploys feed the same dashboard while
                        # remaining filterable.
                        "sessionAttributes": {
                            "applicationName": self.stack_name,
                        },
                    },
                },
            ),
        ],
        destination_bucket=bucket,
        log_group=bucket_deployment_log_group,
    )
    # Defer the slow asset deploy until after the RUM custom resources
    # have succeeded. If RumExtendedMetrics fails (it depends on IAM
    # propagation), the BucketDeployment never starts — saving the most
    # expensive single resource from being repeated on every retry until
    # the cheaper IAM dance settles.
    bucket_deployment.node.add_dependency(rum_extended_metrics)

    # CloudFront cache invalidation, decoupled from BucketDeployment.
    # Defines on_create and on_update only — no on_delete — so CFN simply
    # removes this resource from stack state during teardown without any
    # CloudFront API call to race with the distribution's own deletion.
    # This is the permanent fix for aws/aws-cdk#15891, replacing the
    # BucketDeployment's built-in invalidation hook.
    #
    # CallerReference is gated on the BucketDeployment's content-hashed S3
    # object key. Same assets → same key → CFN sees no change → no invalidation
    # fires (correct: nothing to invalidate). Different assets → different key →
    # CFN fires on_update → invalidation runs. Prevents backend-only deploys from
    # burning the 1000/month free invalidation quota. See README "Design
    # decisions" for the longer write-up.
    #
    # NOTE: CloudFront caps CallerReference at 128 characters, so we cannot fold
    # the api_url (or other identifiers) into it — the content-hashed object key
    # alone is already ~68 chars. If a backend-only API replacement changes
    # config.json's contents, that path is covered because Source.json_data
    # writes config.json as part of the BucketDeployment asset, so its hash —
    # and thus this object key — changes when api_url changes.
    #
    # Rollback caveat (accepted): CloudFront treats a repeated CallerReference
    # as an idempotent replay of the earlier invalidation, so rolling assets
    # back to a previously deployed content hash reuses that deploy's
    # CallerReference and NO new invalidation runs — viewers keep the
    # rolled-back-FROM version cached until TTL expiry. A deploy-unique nonce
    # would fix that but would also invalidate on every backend-only deploy,
    # defeating the quota-saving design. After an asset rollback, run a manual
    # invalidation (CloudFrontDistributionId output) if freshness matters.
    # object_keys is a CDK list-token, not a Python list — use Fn.select.
    cf_invalidation_call = cr.AwsSdkCall(
        service="CloudFront",
        action="createInvalidation",
        parameters={
            "DistributionId": distribution.distribution_id,
            "InvalidationBatch": {
                "Paths": {"Quantity": 1, "Items": ["/*"]},
                "CallerReference": Fn.select(0, bucket_deployment.object_keys),
            },
        },
        physical_resource_id=cr.PhysicalResourceId.of(f"{self.stack_name}-cf-invalidation"),
    )
    invalidate_cf_cache = cr.AwsCustomResource(
        self,
        "InvalidateCloudFrontCache",
        on_create=cf_invalidation_call,
        on_update=cf_invalidation_call,
        policy=cr.AwsCustomResourcePolicy.from_statements(
            [
                iam.PolicyStatement(
                    actions=["cloudfront:CreateInvalidation"],
                    resources=[
                        f"arn:{Stack.of(self).partition}:cloudfront::{Stack.of(self).account}:distribution/{distribution.distribution_id}"
                    ],
                ),
            ]
        ),
        log_group=custom_resource_log_group,
    )
    invalidate_cf_cache.node.add_dependency(bucket_deployment)
    # CDK generates an inline default policy on the AwsCustomResource's
    # auto-created role. Same constraint as the RUM custom resources;
    # apply the same IAMNoInlinePolicy suppressions.
    cf_invalidation_inline_reason = (
        "AwsCustomResource policy is a single least-privilege inline statement scoped to "
        "cloudfront:CreateInvalidation on this stack's distribution ARN — managed-policy "
        "reuse adds nothing"
    )
    acknowledge_rules(
        invalidate_cf_cache,
        [
            {"id": "NIST.800.53.R5-IAMNoInlinePolicy", "reason": cf_invalidation_inline_reason},
            {"id": "HIPAA.Security-IAMNoInlinePolicy", "reason": cf_invalidation_inline_reason},
            {"id": "PCI.DSS.321-IAMNoInlinePolicy", "reason": cf_invalidation_inline_reason},
        ],
    )

    CfnOutput(
        self,
        "CloudFrontDomainName",
        description="CloudFront distribution domain name — use this as your frontend URL",
        value=f"https://{distribution.distribution_domain_name}",
    )
    CfnOutput(
        self,
        "CloudFrontDistributionId",
        description="CloudFront distribution ID — needed for manual cache invalidations",
        value=distribution.distribution_id,
    )
    CfnOutput(
        self,
        "FrontendBucketName",
        description="S3 bucket storing the frontend static assets",
        value=bucket.bucket_name,
    )
    CfnOutput(
        self,
        "RumAppMonitorId",
        description="CloudWatch RUM app monitor ID — used by the browser RUM client",
        value=rum_app_monitor.attr_id,
    )
    CfnOutput(
        self,
        "RumIdentityPoolId",
        description="Cognito Identity Pool ID — used by the browser RUM client for guest credentials",
        value=rum_identity_pool.ref,
    )

    # ── RUM / Cognito cdk-nag suppressions ───────────────────────────────
    # Unauthenticated identities are intentional — browsers have no prior
    # identity and RUM's guest-credentials model is the standard pattern.
    # The role's only permission is rum:PutRumEvents on this monitor.
    acknowledge_rules(
        rum_identity_pool,
        [
            {
                "id": "AwsSolutions-COG7",
                "reason": "RUM requires unauthenticated guest credentials for anonymous browser telemetry",
            },
        ],
    )
    # The guest role has a single least-privilege permission — rum:PutRumEvents
    # on exactly one monitor ARN — tightly bound to this role's one purpose.
    # A managed policy would add indirection without changing the security
    # posture, since the policy is used by nothing else and is scoped to a
    # resource that is itself one-to-one with the role.
    inline_policy_reason = (
        "Single least-privilege inline policy (rum:PutRumEvents on one monitor ARN) "
        "is tightly bound to this role's sole purpose — anonymous browser telemetry upload"
    )
    acknowledge_rules(
        rum_unauth_role,
        [
            {"id": "NIST.800.53.R5-IAMNoInlinePolicy", "reason": inline_policy_reason},
            {"id": "HIPAA.Security-IAMNoInlinePolicy", "reason": inline_policy_reason},
            {"id": "PCI.DSS.321-IAMNoInlinePolicy", "reason": inline_policy_reason},
        ],
    )

    # Explicit CMK log group + singleton nag suppressions for the CDK
    # auto-delete-objects provider Lambda (see create_auto_delete_objects_log_group
    # in nag_utils — it owns both, so there's no per-stack block here).
    create_auto_delete_objects_log_group(self, frontend_encryption_key)

    self._create_athena_glue_resources(access_log_bucket, frontend_encryption_key)

    # ── Per-resource cdk-nag suppressions ──────────────────────────────────
    # All Lambdas in this stack are CDK-managed singletons. Their construct
    # IDs are stable (hashed from CDK's own source) but they are created as
    # stack-level siblings of the construct that requested them, so we look
    # them up with ``try_find_child`` rather than absolute path strings —
    # this keeps the suppression working regardless of whether the stack is
    # at the App root or nested inside a cdk.Stage.
    #
    # Stable singleton IDs:
    #   Custom::CDKBucketDeployment8693BB64968944B69AAFB0CC9EB8756C — BucketDeployment provider
    #   Custom::S3AutoDeleteObjectsCustomResourceProvider — auto-delete provider
    #   AWS679f53fac002430cb0da5b7982bd2287 — AwsCustomResource provider Lambda
    #     (used by RumMetricsDestination, RumExtendedMetrics, InvalidateCloudFrontCache)
    suppress_cdk_singletons(
        self,
        (
            BUCKET_DEPLOYMENT_PROVIDER_ID,
            AWS_CUSTOM_RESOURCE_PROVIDER_ID,
        ),
    )

    # ── Async failure destinations for the CDK-managed provider Lambdas ─────
    # See BackendStack for the full rationale — CFN invokes the providers
    # async, and without on_failure a crashed provider's payload is lost.
    # Both stack-level Function-based singletons get the same treatment: the
    # AwsCustomResource provider AND the BucketDeployment handler. (The S3
    # auto-delete provider is a CustomResourceProvider, not a Function — its
    # async config is not reachable through the L2, which is exactly what
    # its Serverless-LambdaDLQ suppression in CDK_LAMBDA_SUPPRESSIONS
    # documents.) attach_async_failure_destination also emits a CfnOutput of
    # each DLQ URL, so the returned queues are surfaced for operators.
    self.cr_provider_dlq = attach_async_failure_destination(
        self,
        AWS_CUSTOM_RESOURCE_PROVIDER_ID,
        encryption_key=frontend_encryption_key,
        queue_id="AwsCustomResourceProviderDlq",
    )
    self.bucket_deployment_dlq = attach_async_failure_destination(
        self,
        BUCKET_DEPLOYMENT_PROVIDER_ID,
        encryption_key=frontend_encryption_key,
        queue_id="BucketDeploymentProviderDlq",
    )

    self._acknowledge_bucket_deployment_grants(bucket)

    # minimizePolicies restructures the BucketDeployment handler's inline
    # policy into a separate resource under DeployFrontend/CustomResourceHandler.
    deploy_frontend = self.node.try_find_child("DeployFrontend")
    if deploy_frontend is not None:
        suppress_cdk_singletons(deploy_frontend, ("CustomResourceHandler",))

    # ── Stack-level cdk-nag suppressions (genuinely stack-wide) ─────────────
    replication_reason = "S3 replication not needed for sample app — static assets are redeployable"
    versioning_reason = "S3 versioning not needed for sample app — static assets are redeployable via cdk deploy"
    stack_suppressions = [
        ("AwsSolutions-CFR1", "Geo restriction not required for sample app"),
        ("AwsSolutions-CFR4", "Using default CloudFront certificate — no custom domain for sample app"),
        ("NIST.800.53.R5-S3BucketReplicationEnabled", replication_reason),
        ("NIST.800.53.R5-S3BucketVersioningEnabled", versioning_reason),
        ("HIPAA.Security-S3BucketReplicationEnabled", replication_reason),
        ("HIPAA.Security-S3BucketVersioningEnabled", versioning_reason),
        ("PCI.DSS.321-S3BucketReplicationEnabled", replication_reason),
        ("PCI.DSS.321-S3BucketVersioningEnabled", versioning_reason),
    ]
    acknowledge_rules(
        self,
        [{"id": rule, "reason": reason} for rule, reason in stack_suppressions],
    )