Skip to content

Shelfery Implementation Alignment Record (pantryplus)

This record documents the 2026-07-11 evaluation of Shelfery's de facto implementation repository (pantryplus, local C:\Users\Stuart\repos\pantryplus, GitHub brokenlyre/pantryplus) against this repository's governance and venture documentation. Facts about the implementation below are observed evidence read from that repository's docs/ai/ knowledge base and code layout on 2026-07-11; they were not independently re-verified against the running product.

1. What the implementation actually is (observed)

  • A production web application deployed on Vercel: Next.js 16 (App Router, TypeScript), Tailwind CSS v4, shadcn/ui on Base UI, Supabase (Postgres + RLS), Clerk (social-only auth), Anthropic API for AI product identification, @zxing/browser barcode fallback.
  • Delivery model: mobile-first, iPhone-first PWA (Add to Home Screen), not App Store distributed.
  • Shipped capabilities include: multi-user households with admin/member roles and invites; locations/storage areas; scan-first item entry (native BarcodeDetector + AI photo identification + cross-household UPC cache); shopping list with put-away flow; multi-lot expiration tracking with FIFO consumption; used-vs-tossed event capture; AI usage soft caps; a households.plan entitlements seam (free/comped/paid); warm dark mode; kitchen search.
  • Recorded decisions in that repository include: the rename PantryPlus → Shelfery with the shelfery.app domain acquired and a my./test./try. subdomain plan (decided, not yet executed); a long-term goal of paying users with a freemium-subscription strategy document (docs/ai/revenue-ops.md, unapproved strategy); a future Platform Admin Console on its own subdomain.

2. Corrections this evaluation made in this repository

Previously this repository stated Shelfery had shipped nothing and treated platform, AI features, barcode scanning, and household sharing as open questions. Those statements were stale. Updated as part of this task: docs/strategy/venture-brief.md, docs/strategy/assumptions-and-open-questions.md, docs/strategy/roadmap.md, docs/strategy/brand-and-positioning.md, docs/architecture/repository-boundary.md, product/README.md, venture.json, the portfolio registries, and ADR-0010 (proposed).

3. Governance gaps and inconsistencies (with recommendations)

3.1 Process-gate inversion (this repository's gates were bypassed)

The implementation reached production before this repository's Definition gates (approved venture direction → product brief → requirements → build) were ever exercised. This is a factual observation, not a retroactive fault-finding — the implementation predates much of this governance structure. Recommendation: treat the gates as applying from now on (new features/products), and authorize a one-time specification back-fill of ventures/shelfery/product/ from observed product behavior rather than pretending the product is pre-definition. Requires a human decision.

2026-07-27 update: this recommendation is superseded, not acted on as written. Per ADR-0017, this repository does not hold Shelfery's product specification at all — ventures/shelfery/product/ was retired rather than back-filled. Any specification back-fill from observed product behavior, if the Founder still wants one, belongs in the shelfery implementation repository's own docs/ai/, not here.

3.2 No governance linkage in pantryplus

pantryplus/AGENTS.md and docs/ai/ make no reference to Devonshire Digital LLC, this repository, or any company-level policy. Material venture-level decisions (rename, monetization direction, admin-console architecture) are recorded only in pantryplus/docs/ai/decisions.md. Recommendation (pantryplus refactor): add a short "Governance" section to pantryplus/AGENTS.md pointing to this repository as the venture-level authority, and adopt a convention that decisions meeting this repository's "material decision" bar are mirrored here as ADRs or strategy-doc updates.

3.3 Food-claims standard not wired into the implementation repo

This venture's binding-guidance standard (docs/domain/trust-safety-and-food-claims.md) is unknown to pantryplus. Today's shipped features are comparatively low-risk (expiration dates are user-entered, not app-asserted), but several planned features cross directly into the standard's territory: expiry alerts, put-away "typical expiry window" defaults, AI meal ideas from expiring items ("Use It Up"), and any AI-estimated shelf-life or dollar-value figures (R3 Pantry Recap). Recommendation (pantryplus refactor): reference the standard from pantryplus/docs/ai/design-principles.md (or coding-standards) and check the B6/R3/expiry-alert blueprints against sections 1, 3, 4, and 6 before build; route any default-expiry-window feature through the standard's authoritative-source requirement (open question Q5 here).

3.4 Rename decided but unexecuted — two names in circulation

pantryplus decided the Shelfery name and acquired shelfery.app, but the repository, code strings, deployment, and docs still say "PantryPlus" everywhere (including a proposed paid-tier name "PantryPlus Plus" in revenue-ops.md that predates the rename). This repository, meanwhile, already calls the venture Shelfery. Recommendation (pantryplus refactor): execute the App Rename Runbook in pantryplus/docs/ai/implementation-blueprints.md (or explicitly schedule it), and sweep revenue-ops.md for stale "PantryPlus"-derived naming. Until then, readers should treat "PantryPlus" as the working name of the Shelfery product.

3.5 Name-approval status contradiction inside this repository

ventures/shelfery/venture.json says public_name_status: "confirmed", while docs/strategy/brand-and-positioning.md lists the name as "Working name, not confirmed as final." The pantryplus rename decision and domain purchase strengthen the case for "confirmed," but no formal brand approval step has occurred here. Surfaced rather than silently resolved — needs a human call (it is reflected as an open question in the assumptions register).

3.6 Brand conclusions in pantryplus: consistent, still provisional

The implemented brand (warm & earthy; terracotta primary oklch(0.48 0.12 50); sage accent; warm off-white/brown-black backgrounds; "kitchen companion" voice; line-art iconography) matches this repository's provisional direction and is treated as possible but not approved, per the task's framing. No umbrella-brand ("DDLabs"/"Devonside Labs") language appears anywhere in pantryplus — the brand-independence constraint is satisfied. Recommendation: when the formal brand approval step runs, use the implemented tokens as the concrete candidate palette.

3.7 Evidence-labeling drift

pantryplus/docs/ai/revenue-ops.md contains market/behavioral claims (e.g., App Store payment-trust and discoverability assertions) stated as fact without sources. Within pantryplus's own conventions this is acceptable strategy writing; under this repository's evidence principles it would need labeling as assumption/inference. Recommendation: if that strategy is ever mirrored here or used in investor/marketing material, re-label claims per docs/company/research-and-evidence-principles.md first. No pantryplus change required for internal use.

3.8 Minor hygiene items (pantryplus, low priority)

  • Stale GitHub owner reference stuartmdavis/pantryplus vs. actual brokenlyre/pantryplus (already tracked in pantryplus open-questions).
  • R0 Part B analytics events merged but non-functional (Vercel Pro gating) — already tracked there; noted here only because monetization decisions in this repository should not assume analytics data exists yet.
  • No accessibility target documented (tracked in pantryplus open-questions); if this repository later adopts a company accessibility standard, pantryplus inherits it.

4. What was deliberately not done

  • No edits were made to the pantryplus repository.
  • ventures/shelfery/product/ specifications were not back-filled (needs authorization — see §3.1). 2026-07-27: moot — that directory was retired outright per ADR-0017; see the update note in §3.1.
  • ventures/shelfery/AGENTS.md section 6 and docs/ai/guardrails.md still list platform/AI/barcode/household-sharing as "do not treat as decided"; those lists are now partially stale but changing a policy's meaning was out of scope for this task. They should be revised once ADR-0010 is ratified, since identifiable venture documents (this record, the updated strategy docs) now attest to those facts.

5. Addendum — 2026-07-14 spot check

The rest of this record is the original 2026-07-11 evidence snapshot, left as-written per this repository's point-in-time-record convention. This addendum only notes what changed in pantryplus between then and 2026-07-14 (checked against pantryplus commit history):

  • All work in that window was docs-only (a GitHub-owner reference cleanup matching this repo's own 2026-07-13 org-migration cleanup, and one operational note below); no product features shipped in that window beyond what §3 already reflects (commit history through the 2026-07-11 reconciliation already covered B5 Used vs. Tossed, P1 Put Away Groceries, R0 usage caps/analytics, and R1 entitlements — all consistent with what §3's plan entitlements and analytics-blocker notes already describe).
  • §3.8's stale-owner-reference bullet is now resolved in pantryplus (brokenlyre/pantryplus references corrected to the Devonshire-Digital org, 2026-07-13) — no longer an open hygiene item.
  • §3.8's R0 Part B analytics-blocker bullet may be partially stale: on 2026-07-13 pantryplus upgraded its Vercel account to a Pro plan (for an unrelated reason — the repo became private under the org and Hobby no longer covers private-org repos). Pantryplus's own decision record flags this "may also unlock analytics features that were previously blocked on the Hobby plan — worth rechecking." As of this addendum that recheck hasn't happened yet in pantryplus, so this repository's monetization docs should keep treating shipped analytics data as not yet confirmed available, per §3.8's original caution, until pantryplus re-verifies R0 Part B end-to-end.

Report an issue about this page