Back to the Blueprint Audit Download the sample PDFs: founder edition · developer edition
QodeupQodeup Blueprint Audit
App TallyroomDelivered 5 November 2026Audit a-sample-01

SAMPLESample report. The app, its code and every finding are invented to show the format; no real client is described.

App TallyroomDelivered 5 November 2026Audit a-sample-01

Tallyroom works, and two problems need fixing this week.

Any visitor can read every customer's invoices without signing in, and the checkout lets the browser set the price. Both are quick to fix. The rest of the app is in reasonable shape: sign-up, invoicing and payments work end to end, and nothing leaks your secret keys.

What is solid

Reported as clearly as what is broken: secrets, payments and the main flows held under every check that touched them.

  • No secret keys reach the browser, and the git history is clean.
  • Stripe webhooks are verified and cannot be replayed to grant access twice.
  • Sign-up, sign-in and paying an invoice all work at desktop and phone width.
  • Every money amount is stored as whole pence, so totals never drift.

Urgent

2findingsTwo ways in: anyone can read invoices, and the browser sets the price. Close both this week.

1
P0 Urgent

Anyone can read every customer's invoices without signing in

Create an invoice, database Effort: Under a day Fix in: week 1 Check SEC-RLS-01a-sample-01-F-001

Why it matters

The invoices table has no access rules, so a stranger with the public app key can download every invoice, client name and amount. That is a data breach you would have to report.

What to doFix before anything else. Paste the fix prompt into your coding agent, or ask your developer; it is under an hour of work.

The invoices table has no access rules, so a stranger with the public app key can download every invoice, client name and amount. That is a data breach you would have to report.

Evidence

supabase/migrations/20260811_invoices.sql:14cited line highlighted
create table public.invoices (  id uuid primary key default gen_random_uuid(),  owner_id uuid references auth.users not null,
policies/export.sql:88cited line highlighted
-- invoices: row level security DISABLED

Fix

Turn on row-level security for invoices and add policies so each signed-in user reads and writes only rows where owner_id is their own id.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

In Supabase, enable and force row-level security on the public.invoices table. Add select, insert, update and delete policies that allow a row only when owner_id = auth.uid(). Before changing anything, list any page or function that reads invoices without a signed-in user, and tell me; then show the before and after policies and confirm signed-in users still see their own invoices.

2
P0 Urgent

A customer can change the price they pay at checkout

Choose and pay, server Effort: Under a day Fix in: week 1 Check SEC-PRICE-01a-sample-01-F-002

Why it matters

The checkout takes the amount from the browser, so anyone who edits the request can pay one pence for a plan that costs forty pounds. You lose money on every such sign-up.

What to doFix this week, together with the invoices fix.

The checkout takes the amount from the browser, so anyone who edits the request can pay one pence for a plan that costs forty pounds. You lose money on every such sign-up.

Evidence

supabase/functions/create-checkout/index.ts:21cited line highlighted
const { priceAmount, planId } = await req.json()const session = await stripe.checkout.sessions.create({  line_items: [{ price_data: { unit_amount: priceAmount,

Fix

Look up the price on the server from the plan id, and never read an amount from the request. Add an idempotency key to the session call.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

In supabase/functions/create-checkout/index.ts, stop reading priceAmount from the request body. Look up the Stripe price for planId on the server (from a plans table or Stripe price ids), pass it as price, and add an idempotency key built from the user id and plan id. Keep the current checkout flow working for every existing plan.

Serious

2findingsA dead button your customers have already met, and nothing stopping password guessing.

3
P1 Serious

The Delete invoice button does nothing

Delete an invoice, browser Effort: Under a day Fix in: week 1 Check PROD-STUB-01a-sample-01-F-003

Why it matters

Users click it, nothing happens, and the invoice stays. They lose trust and email you instead, which costs you time on every account.

What to doQuick fix; include it in this week's round.

Users click it, nothing happens, and the invoice stays. They lose trust and email you instead, which costs you time on every account.

Evidence

Screenshot: Clicked Delete invoice at 1280 wide: no request sent, no message, the invoice remains.

https://app.tallyroom.example/invoices/demo

Clicked Delete invoice at 1280 wide: no request sent, no message, the invoice remains.

src/pages/Invoice.tsx:142cited line highlighted
<Button variant="destructive" onClick={() => {}}>Delete invoice</Button>

Fix

Wire the button to a delete call with a confirmation step, then return to the invoice list with a message.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

In src/pages/Invoice.tsx, make the Delete invoice button open a confirmation dialog, then delete the invoice through Supabase (respecting the new row-level security), show a short confirmation message, and return to the invoices list. Do not change any other button.

4
P1 Serious

Nothing slows down someone guessing passwords

Sign in, server Effort: One to three days Fix in: week 2 Check SEC-RATE-01a-sample-01-F-004

Why it matters

An attacker can try thousands of passwords against one account without being stopped. Accounts with weak passwords will eventually be taken over.

What to doPlan for next week; your developer should do this one.

An attacker can try thousands of passwords against one account without being stopped. Accounts with weak passwords will eventually be taken over.

Evidence

src/lib/auth.ts:9cited line highlighted
export const signIn = (email, password) =>  supabase.auth.signInWithPassword({ email, password })

Fix

Turn on Supabase's auth rate limits and add a growing cooldown after failed attempts, by IP and account together.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

Add brute-force protection to sign-in: enable Supabase auth rate limits in the dashboard, and add an increasing cooldown after failed attempts, throttled by IP and account together, so nobody can lock a known user out on purpose. Show the user a clear message while the cooldown runs.

Should fix

3findingsQuality and upkeep: errors nobody hears about, a pricing page that breaks on phones, a slow first page.

5
P2 Should fix

When the app breaks, nobody finds out

Across the app: whole app Effort: Under a day Fix in: week 2 Check SHIP-ERRTRACK-01a-sample-01-F-005

Why it matters

Errors that users hit are not recorded anywhere, so you learn about problems only when someone complains, or leaves.

What to doWorth doing next week; it pays for itself the first time something breaks.

Errors that users hit are not recorded anywhere, so you learn about problems only when someone complains, or leaves.

Evidence

src/main.tsx:1cited line highlighted
import React from 'react'import ReactDOM from 'react-dom/client'import App from './App'

Fix

Add an error tracker (Sentry has a free tier) to the front end and the Supabase functions.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

Add Sentry error tracking to this app: initialise it in src/main.tsx for the front end and in each Supabase edge function, using environment variables for the key. Do not change any app behaviour. Tell me how to see a test error arrive.

6
P2 Should fix

The pricing table runs off the screen on a phone

Choose and pay, browser Effort: Under a day Fix in: week 2 Check PROD-MOBILE-01a-sample-01-F-006

Why it matters

On a phone the plans scroll sideways and the price of the middle plan is cut off, so visitors on mobile are less likely to sign up.

What to doInclude in next week's round.

On a phone the plans scroll sideways and the price of the middle plan is cut off, so visitors on mobile are less likely to sign up.

Evidence

Screenshot: At 390 wide the page is 612 pixels wide; the Team plan price is outside the screen.

https://app.tallyroom.example/pricing

At 390 wide the page is 612 pixels wide; the Team plan price is outside the screen.

Fix

Stack the plans vertically below 640 pixels wide.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

On the pricing page, make the plans stack vertically on screens narrower than 640 pixels, with no sideways scroll, and keep the recommended plan first. Check it at 390 wide.

7
P2 Should fix

The first page downloads more code than it needs

Across the app: speed Effort: One to three days Fix in: week 3 Check PERF-CODESPLIT-01a-sample-01-F-007

Why it matters

Visitors wait about four seconds on a phone before the landing page responds, because the whole charting library loads up front. Slow first pages lose sign-ups.

What to doLater; your developer can do this in a morning.

Visitors wait about four seconds on a phone before the landing page responds, because the whole charting library loads up front. Slow first pages lose sign-ups.

Evidence

scans/lighthouse-landing.jsonscanner output
"total-byte-weight": 1.34 MB of JavaScript before first interaction

Fix

Load the charts only on the pages that show them.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

The landing page loads the charting library although it shows no charts. Load the chart components with a dynamic import only on the dashboard and reports pages, and tell me the JavaScript size of the landing page before and after.

Polish

1findingPolish: leftover sample pages from the builder that make the app look unfinished.

8
P3 Polish

Leftover sample pages from the builder are still live

Across the app: tidiness Effort: Under a day Fix in: week 3 Check SHIP-BUILDERLEFTOVER-01a-sample-01-F-008

Why it matters

A demo page with fake customers is reachable at /examples. It looks unfinished to anyone who finds it.

What to doTidy-up; do it whenever convenient.

A demo page with fake customers is reachable at /examples. It looks unfinished to anyone who finds it.

Evidence

https://app.tallyroom.example/examples

Page loads with the builder's demo data.

src/pages/Examples.tsx:1cited line highlighted
// Generated example page

Fix

Delete the examples page and its route.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

Remove the builder's example page (src/pages/Examples.tsx) and its /examples route. Make sure no navigation links point to it.

Make it easier to use

1findingNot defects: one change that would make sending an invoice feel finished, with the reason behind it.

9
P3 Polish

Sending an invoice ends on a toast that disappears

Create an invoice, browser Effort: Under a day Fix in: week 3Based on: A likely costUX law: Peak-End Rule Check PROD-EASE-01a-sample-01-F-009

Why it matters

After Send, a two-second message says Sent and the page goes back to the invoice list. There is no record on the invoice of when it went or to which address, so a user who looks away has to open their sent mail to check, or sends it twice.

What to doDo it when convenient; it is a small change.

What this rests onSending is the moment the app earns its keep. When it ends without a lasting confirmation, users check by hand or send the same invoice again.

After Send, a two-second message says Sent and the page goes back to the invoice list. There is no record on the invoice of when it went or to which address, so a user who looks away has to open their sent mail to check, or sends it twice.

What this rests onSending is the moment the app earns its keep. When it ends without a lasting confirmation, users check by hand or send the same invoice again.

Evidence

https://app.tallyroom.example/invoices/INV-0142

After Send, a toast shows for about two seconds, then the list loads. The invoice page shows no sent date or recipient.

Fix

End the send on a confirmation that stays: the invoice shows Sent, the date and time, and the address it went to, with a way to resend.

Fix promptPaste into Lovable, Cursor, Bolt, Claude Code or any coding agent

When an invoice is sent, keep the user on the invoice and show a lasting confirmation: Sent, the date and time, and the recipient's email, with a Resend button. Store the sent time and recipient on the invoice so the confirmation is still there after a reload. Do not change how the email itself is sent.

Sign up

Desktop: Account created, welcome screen shown.

Phone: Works; the form fits the screen.

Checks: PROD-FLOW-01

Sign up at 1280 wide
Desktop, 1280pxWorks
Sign up at 390 wide
Phone, 390pxWorks

Choose a plan

Desktop: Plans readable side by side.

Phone: Table runs off the screen; see finding F-006.

Checks: PROD-MOBILE-01

Choose a plan at 1280 wide
Desktop, 1280pxWorks
Choose a plan at 390 wide
Phone, 390pxBroken

Create and send an invoice

Desktop: Created and sent in under a minute.

Phone: Works, but the line-item table needs sideways scrolling.

Checks: PROD-FLOW-04

Create and send an invoice at 1280 wide
Desktop, 1280pxWorks
Create and send an invoice at 390 wide
Phone, 390pxWorks with problems

Delete an invoice

Desktop: Button does nothing; see finding F-003.

Phone: Same as desktop.

Checks: PROD-STUB-01

Delete an invoice at 1280 wide
Desktop, 1280pxBroken
Delete an invoice at 390 wide
Phone, 390pxBroken

Every check, one mark each

Security15 of 20 passed
Data7 of 8 passed
Reliability6 of 9 passed
Performance4 of 7 passed
Shipping and monitoring3 of 6 passed
AI featuresNot applicable
Functionality and usability17 of 26 passed
PassedPartlyFailedNot checkedNot applicable

Graded against the Qodeup Check Catalogue v1.4: 76 of 90 checks apply to this app.

Every check, with its question and result

CheckAreaWhat it asksResult
SEC-AUTHZ-01SecurityDoes every server-side write take the caller's identity from the verified session or token, rather than from a user id the client sends in the request body or a hidden form field?Passed
SEC-AUTHZ-02SecurityDoes every endpoint that takes a record id check that the calling account is actually allowed to reach that specific record, not just that the caller is logged in?Passed
SEC-CORS-01SecurityDoes every API that requires authentication answer CORS with an explicit origin list, rather than a wildcard?Passed
SEC-CSP-01SecurityDoes the app send a Content-Security-Policy header, rather than relying only on the framework's default HTML escaping to stop injected scripts?Partly
SEC-DEPS-01SecurityDoes the app's dependency tree carry any known-exploited, critical or high-severity vulnerability in a package the running app actually uses on a reachable code path?Passed
SEC-KEYS-01SecurityIs the service-role key, or any other database or admin secret key (a private API token that isn't a model-provider key), kept out of every file the browser bundle can reach, including public env var prefixes?Passed
SEC-KEYS-02SecurityDoes the app refuse to start, rather than fall back to a hardcoded or generated default, when a required secret key is missing from the environment?Passed
SEC-RATE-01SecurityDoes the login endpoint specifically lock out or slow down repeated failed attempts, rather than relying only on a generic site-wide rate limit?Failed
SEC-RATE-02SecurityAre public, unauthenticated write endpoints (contact forms, sign-up, password reset, comment posting) rate limited and returning 429 once the limit is hit?Passed
SEC-REDIRECT-01SecurityIs the redirect target after login, logout or an auth callback checked against an allowlist of the app's own paths, rather than built directly from a client-supplied parameter?Passed
SEC-RLS-01SecurityDoes every table in an exposed schema have row-level security turned on and forced, with a policy for each operation the app actually performs?Failed
SEC-RLS-02SecurityDo the row-level security policies scope rows to the record's owner, rather than granting access to anyone who is simply signed in?Partly
SEC-SECRET-01SecurityIs the working tree and git history free of committed secrets, keys and database dumps?Passed
SEC-SESSION-01SecurityAre session cookies set with HttpOnly, Secure and an appropriate SameSite value?Passed
SEC-CALLBACK-01SecurityDoes the sign-in callback (OAuth, or an emailed sign-in link) check the provider's code and state, or the link's token, on the server before it creates a session, instead of trusting whatever the browser sends back?Passed
SEC-CSRF-01SecurityAre state-changing requests authenticated by a cookie protected by a CSRF token, a SameSite setting that blocks cross-site posts, or an Origin check, rather than the cookie alone?Passed
SEC-PRICE-01SecurityDoes the server decide the amount, currency and product for every charge, with nothing the browser sends able to change the price, and does charge creation carry an idempotency key so a retry cannot charge twice?Failed
SEC-SESSION-02SecurityDoes a session end on logout, expire on its own, and get revoked when the password changes, rather than staying valid indefinitely on an old token?Passed
SEC-STORAGE-01SecurityAre storage buckets private by default, with upload and read policies scoped so a user only reaches their own files, and public buckets used only where the content is genuinely meant to be public?Not applicable
SEC-VIEWS-01SecurityDo database views and SECURITY DEFINER functions carry a stated reason and their own access check, rather than silently running with wider rights and bypassing row-level security underneath?Passed
SEC-WEBHOOK-01SecurityDoes every server function correctly classify its caller (a signed-in user, a provider webhook, or a scheduled job) and apply the matching check, instead of one shared authentication path for all three?Passed
DATA-BACKUP-01DataAre automated backups running, and has a restore actually been performed and timed at least once?Passed
DATA-MIGRATION-01DataAre all schema changes made through versioned migration files, rather than edited directly in a dashboard?Passed
DATA-PII-01DataIs the personal data this app collects classified somewhere written down: what is collected and why?Partly
DATA-RETENTION-01DataIs there a stated retention and deletion policy, and can a delete request actually be honoured against the database, not just the UI?Passed
DATA-TIME-01DataIs every timestamp stored in UTC and converted to the viewer's local time only for display?Passed
DATA-TXN-01DataDo multi-step writes that must all succeed together (checkout, account deletion, a role change) run inside one transaction or RPC call, rather than as several sequential client calls?Passed
DATA-CONFIG-01DataIs configuration that lives only in a dashboard (auth providers, redirect URLs, custom SMTP, storage limits, scheduled job schedules) written down somewhere the next session or audit can see?Passed
DATA-MONEY-01DataIs every monetary amount stored as an integer in the currency's smallest unit, rather than as a float or decimal that gets formatted at storage time?Passed
REL-ENV-01ReliabilityDoes a complete .env.example exist, and does every key in it actually exist in the production deployment target, not just locally?Passed
REL-ENVSPLIT-01ReliabilityDo staging and production use separate databases and separate provider keys, rather than sharing one?Passed
REL-ERRHANDLE-01ReliabilityWhere a failure is caught, is it logged and surfaced with a safe message, or is the catch block empty and the failure disappears?Failed
REL-KILLSWITCH-01ReliabilityCan any feature that spends money, calls an AI model, or writes to the database be turned off instantly without a code deploy?Passed
REL-RETRY-01ReliabilityDo transient failures on safe-to-repeat outbound calls get retried with backoff, instead of failing the whole request on the first blip?Partly
REL-TIMEOUT-01ReliabilityDoes every outbound call the app makes (fetch, an API client, a database driver, a third-party SDK) carry an explicit timeout?Partly
REL-BGJOB-01ReliabilityDoes slow work (sending an email, calling a third-party API, processing an upload) run inline in the request, or is it handed off to a queue so the user isn't stuck waiting?Not applicable
REL-CIRCUIT-01ReliabilityWhen a critical upstream (payment provider, email sender, the app's own backend service) goes down, does the app stop hammering it, or does every page load keep firing requests at a dead service?Not applicable
REL-CONNPOOL-01ReliabilityIs the app's total database connection use, counted across every concurrent build, deploy preview and serverless worker, budgeted against the pooler's project-wide cap?Passed
REL-RACE-01ReliabilityWhere two requests can update the same record at once (a stock count, a role, a balance, or an 'at least one must remain' rule like the last admin or last owner), is the update atomic and run inside one locked transaction, rather than a read followed by a separate write?Passed
REL-WEBHOOK-01ReliabilityDoes every inbound webhook verify the sender's signature against the raw body, and does replaying the same event twice change nothing the second time?Passed
PERF-CACHE-01PerformanceAre expensive, repeated reads cached with an explicit TTL, and does a write to the underlying data actually invalidate the cached version?Partly
PERF-CODESPLIT-01PerformanceDoes the first screen ship only the JavaScript it needs, or is a heavy component (a chart library, a rich editor, an admin-only feature) bundled into every visitor's initial load?Failed
PERF-COMPRESS-01PerformanceAre responses and static assets actually served compressed and with correct cache headers, or is that left to whatever the platform happens to default to?Passed
PERF-IMAGE-01PerformanceAre images served in a modern format, sized for their container, and lazy-loaded below the fold?Partly
PERF-PAGINATE-01PerformanceDo list endpoints and pages (a blog index, a product catalog, a search result set) load a bounded page of results, or do they return everything at once?Passed
PERF-INDEX-01PerformanceAre the columns the app actually filters, sorts, or joins on covered by an index?Passed
PERF-NQUERY-01PerformanceWhen the app renders a list with related data (posts with authors, orders with line items), does it fetch the related data in one batched query, or does it fire one query per row?Passed
SHIP-BUILDERLEFTOVER-01Shipping and monitoringDoes the repo carry leftover scaffolding, sample data, or duplicate instruction files from the AI builder that made it (a demo route, seed data still live, conflicting CLAUDE.md/cursorrules/AGENTS.md files that disagree with each other)?Failed
SHIP-CI-01Shipping and monitoringDoes the gate before deploy run the build, a type check, lint and the test suite, and block the deploy when any of them fails?Passed
SHIP-ERRTRACK-01Shipping and monitoringDo exceptions from every part of the app (frontend, backend, background jobs) land in one place with user and request context, or does the operator find out about bugs only when a user reports one?Failed
SHIP-LOCKFILE-01Shipping and monitoringIs a lockfile committed and used, so the exact dependency versions installed locally match what the build server installs?Passed
SHIP-LOGS-01Shipping and monitoringAre logs structured (JSON, with a request id) so a specific failed request can be traced, or is it free-form text nobody can search?Passed
SHIP-TEST-01Shipping and monitoringDo the app's critical paths (checkout, auth, contact/signup) have automated tests that run on every change, or is verification manual only?Partly
SHIP-DEPLOYSURFACE-01Shipping and monitoringWhen the app has more than one deployable surface (a web app, a separate admin backend, database migrations, scheduled jobs), is it documented which of them a push actually redeploys automatically?Not applicable
AI-DATA-01AI featuresIs what user or tenant data reaches the model provider known and kept to the minimum needed, is the provider's data retention setting stated, and does retrieval return only the caller's own tenant's documents?Not applicable
AI-EVAL-01AI featuresIs there an eval set the AI feature's prompt or model gets checked against, or is 'it looked better' the entire test suite for every prompt change?Not applicable
AI-INJECT-01AI featuresEverywhere untrusted text reaches the model (a user message, an uploaded file, a fetched page, another user's record), what stops an instruction hidden in that text from being followed?Not applicable
AI-KEY-01AI featuresDoes the AI provider's API key ever reach the browser, in client-side code, a public env var, or a network response?Not applicable
AI-MODELPIN-01AI featuresIs the model version pinned to a specific identifier, or does the code point at a floating alias that can change underneath the app?Not applicable
AI-OUTPUTCHECK-01AI featuresBefore the model's output is saved, shown, or acted on, does anything check its shape and content, or does it flow straight through?Not applicable
AI-RATELIMIT-01AI featuresDoes a paid AI route limit how often a single user can call it, separately from the spend cap, so one user can't monopolise it or trigger abuse patterns even while staying under the money ceiling?Not applicable
AI-SPENDCAP-01AI featuresDoes the AI feature have a per-user and a global spend cap that hard-stops the calls, plus a billing alert on the provider account, or can it run up an unbounded bill?Not applicable
PROD-A11Y-01Functionality and usabilityCan the sign-up or main feature flow be completed using only the keyboard, with no mouse?Passed
PROD-A11Y-02Functionality and usabilityDoes every input on the sign-up, sign-in, and main data-entry forms have a label a screen reader can announce?Passed
PROD-A11Y-03Functionality and usabilityDoes the body text and the primary call-to-action button meet at least 4.5:1 contrast against their background (3:1 for large text), as rendered on the page?Partly
PROD-A11Y-04Functionality and usabilityIs there a visible focus indicator on every interactive element when navigating by keyboard?Passed
PROD-A11Y-05Functionality and usabilityWhen text is enlarged to 200 percent, can the landing page, the sign-up form and the main feature screen still be read and used, with nothing cut off or overlapping?Passed
PROD-A11Y-06Functionality and usabilityAt 320 CSS pixels wide, does the content reflow into one column, so nothing needs sideways scrolling to read or use?Failed
PROD-EASE-01Functionality and usabilityIs the core flow easy to use as well as working, judged against the UX laws index, with a cost or a standard behind every finding?Partly
PROD-DECEPTIVE-01Functionality and usabilityDoes the product work against the person using it anywhere, such as a box ticked before they see it, a cost that appears only at the last step, a decline written to shame, cancelling harder than signing up, or urgency that is not true?Passed
PROD-CONSOLE-01Functionality and usabilityDoes the browser console stay free of errors (not warnings) on the landing page, the sign-up or sign-in screen, and the main feature screen?Passed
PROD-FLOW-04Functionality and usabilityCan the app's one main feature, the thing the product exists to do, be completed end to end at desktop and phone width?Partly
PROD-MOBILE-01Functionality and usabilityDo the landing page, the main feature screen, and any form pages hold their layout at phone width with no sideways scroll?Failed
PROD-MOBILE-02Functionality and usabilityAre the primary tap targets (buttons, nav items, form controls) at least 24 by 24 CSS pixels, or spaced as the WCAG 2.2 target-size rule allows, at 390px phone width?Failed
PROD-STATE-01Functionality and usabilityDoes every list, dashboard, or search result screen show a real empty state when there is nothing to show, instead of a blank area or a frozen loader?Partly
PROD-STATE-02Functionality and usabilityDoes the app show a visible loading indicator during every action that takes more than an instant, rather than leaving the screen looking frozen or unresponsive?Passed
PROD-STATE-03Functionality and usabilityWhen a request fails, does the user see a plain-language message that says what happened and what to do next, instead of a blank screen, a stuck spinner, or a raw error?Passed
PROD-STUB-01Functionality and usabilityDo the buttons and links on the app's primary screens actually do something, rather than being visual stubs left over from the build?Failed
PROD-STUB-02Functionality and usabilityDo the links in the footer and any secondary navigation, not the primary header navigation checked in PROD-STUB-01, all resolve to a real page rather than a 404 or a placeholder?Passed
PROD-FLOW-01Functionality and usabilityCan a new visitor complete sign-up end to end, in a real browser, at desktop and phone width?Passed
PROD-FLOW-02Functionality and usabilityCan a returning user sign in end to end, in a real browser, at desktop and phone width?Passed
PROD-FLOW-03Functionality and usabilityCan a test purchase be completed end to end, from the pricing or product page through to a confirmed paid state, with no dead end?Passed
PROD-FLOW-05Functionality and usabilityWhen a user deletes their own account, is their personal data actually deleted or anonymised in the database, not only the account marked deactivated?Passed
PROD-FORM-01Functionality and usabilityDo the app's forms validate input inline and block submission with a clear message, rather than accepting bad data or failing only after a full round trip?Passed
PROD-FORM-02Functionality and usabilityWhen a form submission fails, does the user's already-typed input stay in place, or does the form clear and force them to start over?Passed
PROD-PAY-02Functionality and usabilityWhen a test payment is declined, does the app tell the user clearly and let them retry, rather than leaving them stuck with no explanation or silently charging nothing while showing success?Passed
PROD-SEO-01Functionality and usabilityDoes every public page carry a unique title, a unique meta description, and a canonical URL, and do internal-only pages carry noindex?Not applicable
PROD-SEO-02Functionality and usabilityWhen the homepage and a content page are shared on WhatsApp, LinkedIn, or Twitter/X, does the link show a real title, description, and image?Partly
PROD-SEO-03Functionality and usabilityDo sitemap.xml and robots.txt exist at the live domain, load without error, and stay in sync as content is added?Not applicable
PROD-SESSION-01Functionality and usabilityDoes refreshing the browser mid-flow, or closing and reopening the tab, keep the user signed in and their in-progress work intact?Passed

Auditor's statement

Qodeup examined Tallyroom's code, database rules and live app against the Qodeup Check Catalogue v1.4.

Every finding was found by the audit crew and challenged by a verifier against its evidence. Every urgent and serious finding was reviewed again, blind, by an independent second model. A senior engineer read the report and signed it.

Orezi Emeotu, founder, Qodeup AfricaSigned 5 November 2026Report version 1, audit a-sample-01