Review

Code review & architecture assessment

For an engineering reviewer doing a BC production-standards sniff test. This page doesn't pre-form a verdict — it gives you what you need to form your own, including the specific list of what isn't verified yet.

How this was built

Built primarily with Claude Code (Anthropic's agentic coding tool), under direct human direction — the architecture, specification, and review decisions are the operator's; the AI generates code against that spec. The output is conventional TypeScript, Astro, and React — standard, maintainable code, not a proprietary format tied to the tool that helped write it. Quality claims here aren't self-assessed: every one is backed by an automated, regenerating verification ledger, not a status doc — see Verified Throughput. For a structured engineering review, see Code Review.

Findings at a glance

G4-verified

202 / 218

Failing

Known limitations

Epics affected

Full partition and methodology: Verified Throughput

Repo access (gate zero)

Nothing below substitutes for opening the source. The repository is private (github.com/nino-chavez/bc-subscriptions) — real read access for a reviewer is an access grant the operator needs to arrange directly, not something this portal can self-serve. Ask for it before your sync; everything else on this page assumes you have it.

Orientation map

Where the money path lives, and which directories are production code vs. design reference.

Path	What it is	Status
apps/api	Cloudflare Workers — the charge/subscribe/dunning/cancel money path, BC platform integration (webhooks, stored-instruments vault)	production
apps/admin	Merchant admin — React + BigDesign (BC's own design system)	production
apps/storefront-svelte, apps/storefront-catalyst	Two storefront integrations — Svelte (live Kibble & Co. demo) and Catalyst (BC's headless React stack)	production
apps/portal	This site — Astro, internal/stakeholder communication, not part of the shipped product	internal tooling
prototype/	Design exploration — read for intent, not a code-quality sample; not shipped	not production

Run it yourself

Independent, reproducible output — re-run these rather than take a number from this page.

npm run test # full test suite, turbo-orchestrated across workspaces

npm run typecheck # TypeScript across all workspaces

npm run test:scenarios # end-to-end behavioral scenarios (the G4 evidence above)

Security scanning (gitleaks, npm audit, Dependabot) already runs continuously — counts at /review.

How status is graded here

"Compliant" means present, not done. These statuses are derived from artifact presence (a file exists, a check passes a structural test) — they are not the same as a passing, end-to-end behavioral scenario. For the verified-working tier, see Verified Throughput.

The rubric below is project-authored, not an external standard. These rows are scored against a capability catalogue this project wrote for itself — useful as a self-check, not independent certification. Treat it as a starting point for your own review, not as proof of conformance.

Known limitations — 16 acceptance criteria not yet G4-verified

Named, not implied. Each row is an acceptance criterion without a passing automated scenario yet — grouped by epic so the gap is locatable, not just countable.

Epic 8 — 4 open

no-gate
US-8.2
Widget on cart
no-gate
US-8.3
Headless SDK
no-gate
US-8.4
Widget theming
no-gate
US-8.6
Pre-purchase education panel

Epic 17 — 1 open

no-gate
US-17.4
Portal theming

Epic 18 — 1 open

no-gate
US-18.11
Click-to-cancel — same-medium cancellation parity

Epic 20 — 1 open

no-gate
US-20.7
Churn-flow obstruction guard (save-attempt limit)

Epic 22 — 1 open

no-gate
US-22.2
Customer lookup from BC App Extension

Epic 23 — 1 open

no-gate
US-23.18
Auto-renewal reminder notifications

Epic 24 — 1 open

no-gate
US-24.11
Catalyst storefront + Buyer Portal + subscription widget composite PDP behaviour

Epic 25 — 2 open

no-gate
US-25.2
Subscribe-and-save auto-promotion
no-gate
US-25.11
Cycle-discount ladder — admin propagation controls

Epic 27 — 1 open

unverified
US-27.4
Headless SDK (TypeScript)

Epic 28 — 3 open

no-gate
US-28.5
SOC 2 Type II readiness
no-gate
US-28.6
Data residency
no-gate
US-28.7
Auto-renewal disclosure policy & consent-record retention

Architecture & platform-fit comparison

The data model, state machine, and BC platform integration seams are rendered as data, not pre-argued, at/architecture. Whether this is a good structural fit against how Shopify made subscriptions a first-class platform citizen is the comparison you're being asked to make — this page deliberately doesn't pre-conclude it for you.

Infra & scale, if brought in-house

Today's footprint is Cloudflare Workers + Pages (Phase 1). A native, in-house path is sized inDelivery fork andDelivery shim path — both ADR-grounded design documents, not load-tested numbers. Be explicit in your own assessment about which figures are decided architecture vs. still-theoretical estimates.

Bringing in a new team?

If the next step is onboarding a team with zero prior context (e.g. a vendor handoff), theteam handoff packet covers the same orientation map and known limitations above, plus a stack map, a scoping-question checklist, and a day-one plan.