Source coverage

This map records the meaningful source responsibilities covered by the website. It joins implementation, tests, canonical reader home, evidence boundary, and editorial review. It covers responsibilities rather than every YAML key. The explicit publication list is site/config/content-map.json; source documentation outside that list remains available as repository references.

The implementation baseline is the public devSatym/resilience-gate repository at 2420ff4c2a88b1fa4a19e413cb0752e271480081, read from the ignored .source/resilience-gate checkout. The standalone private documentation repository owns the reader articles and site engineering files. Source links below use logical paths resolved to their correct repository; the upstream root showcase is linked explicitly to avoid confusing it with this repository’s maintenance README.

Live observations retain their original 1–3 October 2026 identities and windows. “Source reviewed” below means implementation and applicable tests were read for the explanation; it does not claim another live campaign or independent public access to private evidence. Prior monorepo website tests remain under migration-evidence/monorepo-audit/; repository-separation results remain in docs/website/audit/migration/. Cloudflare local results and publication/HTTPS status are separately recorded under docs/website/audit/cloudflare/.

Coverage inventory

Subsystem / responsibility	Canonical reader home	Source	Test contract / recorded boundary	Review and discrepancy
Purpose, topology, reader journeys	Overview, tour, paths	Upstream root showcase, architecture	Verification snapshot	Source reviewed; upstream showcase unchanged. Snapshot is historical.
Application route/data behavior	Application design, HTTP reference	main.py	Request tests, redirect tests	Source reviewed; HTTP behavior derived from routes, not invented endpoints.
Health/startup and dependency recovery	Health, dependency boundaries	main.py, deployment probes	Health, recovery	Source reviewed; Redis required until first successful readiness, PostgreSQL throughout lifetime.
Cache fallback and Redis fault	Redis workflow	main.py, Redis fault	Cache tests, Redis scoring, public card	Source reviewed; ordinary cache misses alone do not prove outage.
Persistence failure/recovery	PostgreSQL workflow, paid request	main.py, PostgreSQL fault	Database tests, PostgreSQL scoring	Source reviewed; recovery does not establish backup restoration/HA.
x402 verification, settlement, replay	Paid request, application design	payment.py, payment contract	Paid requests, replay, private historical paid smoke	Source reviewed; settlement and persistence are not atomic; invalid fields in a facilitator object map to 402 while malformed JSON/non-object bodies map to 503; older prose broadly promises 5xx; public payment identities redacted.
Signer/API/Permit2 bootstrap	Signer workflow, secrets	main.py, permit2.py, signer deployment	API, Permit2, signer scoring	Source reviewed; signer loss can prevent useful paid client traffic while app health remains good.
App/signer/gate container contracts	Delivery design, source validation	App Dockerfile, signer Dockerfile, gate Dockerfile	Container tests, CI contracts	Source reviewed; runtime identities distinct from app source.
Source CI	Source validation	validate workflow, validate.sh, pytest.ini	CI contracts; historical report counts separate	Source reviewed; credential-free may download dependencies; integration check selected separately.
Image publication/OIDC/Cosign	Artifact trust, release journey	app publication, signer publication, gate publication	CI tests, reviewed screenshot 03	Source reviewed; CI verifies after push, no independent Kargo/admission signature enforcement.
Terraform cluster/network/registry	Delivery design, lab lifecycle	GKE, network, registry, variables	Infrastructure tests, October 3 fixed two-node snapshot	Source reviewed; shared zonal cluster, no region/fleet/HA claim.
IAM, CI identity, workload identity	Secrets and identity	GitHub OIDC, IAM, Kargo identity, ESO	Infra contracts, bootstrap secrets tests	Source reviewed; mechanism does not establish organizational audit/rotation policy.
Phased bootstrap/public rendering	Lab lifecycle, configuration	bootstrap, renderer, phases 00–06	Bootstrap, render, Terraform phase	Source reviewed; phase 06 registers contracts, does not promote/run load/chaos.
Read-only verifier and guarded teardown	Lab lifecycle, commands	07-verify, lab-ops	Verifier tests, lab operations	Source reviewed; destroy-plan can read remote state; no teardown claimed.
Helm environments/dependency resources	Configuration, integrated design	chart values, dev, staging, prod	Rendering, secrets	Source reviewed; replica differences do not make stateful stores highly available.
External Secret references/materialization	Secrets flow	store, signer secret, loadgen secret, chart secret	Helm secrets, final 9/9 Ready summary	Source reviewed; website never reads Secret values or ignored config.
Argo CD root/ApplicationSet/ownership	Controller ownership, delivery	root app, AppProject, ApplicationSet	Promotion contracts, reviewed fresh GitOps views	Source reviewed; reconciles rendered output, distinct from release selection.
Kargo Freight discovery and rendering	Release journey, promotion	Warehouse, Project, Stages	Warehouse, staging, prod	Source reviewed; dev automatic, staging/prod manual; manual Freight approval is separate override.
Gate workflow/fault selectors/RBAC	Bounded experiments, gate lifecycle	workflow, RBAC, fault manifests	Selectors, workflow	Source reviewed; planned fault must have applied timestamps; selected sequential pod faults only.
Gate preflight, load, locking, cleanup	Gate lifecycle, troubleshooting	orchestrator	Orchestrator, cleanup, retained direct boundary cases	Source reviewed; pass requires paid preflight and cleanup; quiet counters insufficient for continuous signer-fault traffic. verify_absent treats any failed get as absence, including possible API/read failures.
Prometheus scoring/no-data rules	Measurements, metrics/scoring, score→promotion	scorer, scorecard schema	Queries, no data, scorecard	Source reviewed; or vector(0) intentionally exists for selected expressions; no blanket absence claim.
Dashboards, scraping, Loki/Alloy, annotations	Gate/observability design, gallery	Observability values, ServiceMonitor, Alloy, annotations	Observability, annotation tests	Source reviewed; lab sizing, no monitoring HA/retention/DR claim.
Paid staging and unpaid baseline load	Load testing, formats	loadgen.js, runner, baseline scorer	Load contracts, runner, baseline score	Source reviewed; older runbook omits baseline path; baseline and chaos have distinct formats.
Local Compose/unpaid smoke	Local development	Compose, smoke	Local recovery, health/cache contracts	Source reviewed; maintained smoke redirects before Redis loss, only health checks during loss.
Live request and sanitized evidence tools	Promotion, publish	request runner, collector, utils, metadata schema	Tooling tests	Source reviewed; accepted request/schema-valid supplied status is not verdict. chart_revision is rendered output.
Public evidence/history/availability	Recorded release, historical cases, interpretation	report, evidence index, selected public metadata/log/scorecards	Oct 1–2 public subset; Oct 3 private 36-file/two-record bundles	Source reviewed; privately retained raw records are unavailable to public readers.
Reviewed screenshot inventory	Gallery, recorded release	manifest, accepted PNG files, validator	34 reviewed captures, recorded SHA-256 audit at Oct 3 17:26:01Z	Source reviewed; screenshot 26 predates fresh run; historical dashboards keep their own windows.
Integrated design and non-claims	Integrated design, limitations	design contracts, known limitations	Source/test and named evidence boundaries above	Source reviewed; no underlying application/platform capability added for website.
Static website mapping/build/Git-integrated Pages maintenance	Website design, requirements, audit, publication	site package, content map, upstream pin, preparation script, Pages local configuration, maintenance	Historical monorepo, repository-separation, and Cloudflare records distinguished in audit report	Static hosting adds no platform capability; configured/local/remote status and source identities remain separate.

Editorial discrepancy decisions

The website adapts existing sources without changing operational behavior. It makes the following distinctions explicit:

1. Recorded versus current: the final report’s platform and application sources differ. Historical test counts and resource snapshots are attributed to their recorded commits; a later documentation build does not repeat the campaign.
2. Chart source versus rendered output: Freight’s chart-source commit is distinct from chart_revision in evidence metadata, which means observed rendered branch. Both remain visible in the release case.
3. Trust versus identity: digest pinning preserves content; CI Cosign verification is not independent Kargo or admission enforcement. The pre-sign push/discovery gap is retained as a limit.
4. Procedure versus proof: request acceptance, static rendering, a healthy target, operator-supplied pass status, and a screenshot each have narrower meaning than a completed gate plus cleanup.
5. Public versus private: retained metadata can index unpublished extracts. Selected historical scorecards are public examples; fresh October 3 raw scorecards remain private.
6. Old runbook versus current command: the source load runner includes an unpaid dev baseline path that the older staging runbook does not cover. The operator page documents both and links the implementation.
7. Smoke versus recovery test: the local smoke checks a redirect before Redis loss and health during loss. Cache/recovery tests separately exercise fallback; the website does not overstate the smoke.

The publication boundary excludes ignored config.env, wallet material, raw diagnostics, private archive paths, Terraform state/plans, controller-owned rendered branch outputs, live dashboards, and external runtime content fetching. Operational guides describe existing guarded commands; implementation of the static website performs no live lab operations.