Quality.Noted Help

Quick Start Guide

Quality.Noted runs a phased, release-readiness QA audit of a macOS app's source code. It drives the Claude Code CLI headlessly in your project folder, on your own Claude plan, and turns the output into a readiness score, severity-ranked findings that are tracked across runs, and a manual test plan.

Requirements

• macOS 14+ on Apple silicon.
• The claude CLI on your PATH. Install it with curl -fsSL https://claude.ai/install.sh | bash or brew install --cask claude-code, then verify with claude --version. The Claude desktop app's Code tab is the same Claude Code, but it doesn't always install the CLI. If the binary lives somewhere unusual, set its path in Settings.
• Authenticated: claude auth login. Usage comes from your Claude subscription or API key.
• A project that builds. On Standard and Thorough, the app is built from scratch with your Swift/Xcode toolchain before Phase 1.

What happens to your project

• Execution: each phase is a separate claude -p --dangerously-skip-permissions session with your project as the working directory. It starts in your project but runs as you, so it can read, change or run anything your macOS account can; it's instructed to write only to QualityNoted-QA/. Text in the project (comments, docs) reaches the agent too, so only audit projects you trust. Commit before you audit.
• Briefing: your answers become a briefing written to a delimited section of the project's CLAUDE.md. It's added when an audit starts and removed when it ends; content outside the markers is preserved.
• Build and test: the pre-audit build goes to its own cache folder and doesn't touch your DerivedData. For Swift packages, your tests run too. Xcode test targets are not run.
• Proof programs: to confirm a finding, Claude may compile and run small test programs. If one triggers an unfamiliar Dock icon or alert, that's the reason.
• Output goes to QualityNoted-QA/ in the project:
  • Results/Results_YYYY-MM-DD_HHMM/: PhaseN_Findings.md and .json, RunUsage.json, and the walkthrough files.
  • Rechecks/Recheck_YYYY-MM-DD_HHMM/: one folder per re-check.
  • FindingsLedger.json: every finding with a stable ID, tracked across runs.
• Network: Quality.Noted's only network calls activate and check your license with Polar (when the app opens, at most once a day) and check GitHub for updates (no personal data). Claude Code sends what it reads to Anthropic under your account.

macOS permission prompts

• Files & Folders: appears only if the project is in Desktop, Documents, Downloads, iCloud Drive, or on a removable or network volume. It's asked once per location. If you denied it, change it in System Settings ▸ Privacy & Security ▸ Files & Folders.
• Automation (Terminal): appears only when you click Sign In to Claude Code, which runs claude auth login in Terminal.
• Notifications: appears only if you enable the finish notification.
• Full Disk Access: never required.

Depth, model and effort

• Times were measured on small apps. A Thorough run with Opus on a large codebase can approach two hours.
• Depth sets scope; effort sets reasoning per item. They're independent: Thorough with Low effort still reads every file, just less carefully.
• Phases:
  1. Static Audit: bugs, crashes, security, App Store risk.
  2. Functional QA: flows, edge cases, accessibility, error handling.
  3. Release Cert: readiness for your distribution channel; sets the score.
  4. Manual Checklist.
  5. Tester Walkthrough.

Your first audit

1. ⌘N New assessment. Set the distribution channel (App Store, Homebrew or direct download), the tool that built the app, and the project folder.
2. Review & generate. For a baseline, use Standard · Sonnet · High. Then click Generate & Run Audit.
3. Let it run. You can pause between phases. A phase interrupted mid-way, by Pause or by hitting a usage limit, restarts from the beginning when you resume. Runs are serialized: one at a time, with the rest queued.
4. Read the results.
   • Readiness score: 85–100 is ready, 70–84 close, 55–69 needs work, below 55 not ready.
   • Severity: Critical blocks release. High should be fixed before you ship.
   • Confidence: each finding is tagged Verified (a test program proved it, and the evidence is saved), Proven (the source settles it) or Suspected (verify before acting on it).
   • Files Read: shows actual coverage, for example 101/101.
5. Fix. Copy Action Items Prompt copies every Critical, High and Medium item, worst first. Paste it into a Claude Code session in the project. Each finding also has its own copy button.
6. Re-check (⇧⌘R). A targeted pass over three things:
   • open Critical, High and Medium findings;
   • open Low findings in files changed since the last check;
   • those changed files, scanned for regressions. The scorecard shows Fixed, Still there, New problems and Couldn't re-check. Copy Remaining Action Items Prompt copies the Critical, High and Medium problems still left; Low findings are left out.

Free and full version

Any model and effort in both. The full version is a one-time purchase (early-bird US$14.99, later US$19.99, plus tax where it applies), activated in Settings ▸ License on up to 2 Macs.

Testing approach

The audit follows the seven ISTQB principles of software testing:

1. Testing shows defects are present, not that none exist. Findings carry a confidence tag, Files Read shows coverage, and a ready verdict says "in what was checked".
2. Exhaustive testing is impossible. Depth sets the scope, and every phase weights its effort by risk toward the features you mark as central.
3. Test early. Quick runs and Re-check are cheap enough to use during development, not only before release.
4. Defects cluster. Insights ▸ Defect hotspots ranks the files that hold the most findings.
5. Repeating the same tests stops finding bugs. There are no fixed test scripts. Each audit explores the code fresh, and Re-check scans changed files for new problems.
6. Testing depends on context. Release Cert follows your distribution channel, and your answers about the app shape every phase.
7. Bug-free isn't the same as right for users. The Tester Walkthrough covers what only a person can judge. Validating real users' needs is still up to you.

Tips

• Audit early, not only before release. A Quick run after each major feature catches problems while they're cheap to fix.
• An audit finds defects; it can't prove there are none. "Ready to ship" means a score of 85 or more and no open Critical or High in what was checked. Read Files Read for coverage, and do the Tester Walkthrough for what code review can't judge.
• Use Re-check after each round of fixes. Re-run in full only after significant changes, or with Thorough before a release.
• Stay on one model when comparing runs, so score changes reflect the code and not the model.
• Suggestion: leave Built artifact empty while you're developing, since a development build is expected to fail signing checks. Before you ship, run one last audit with Built artifact pointing at your signed release build and the release-build switch on, to check signing, notarization and Gatekeeper directly.
• Standard and Thorough runs include a Tester Walkthrough of what only a person can verify. Claude Tests What It Can… hands the automatable checks to a Claude Code session.
• Built with another AI tool, or by hand? Choose that option during setup to get prompt files you can paste into any assistant instead of running Claude Code.
• Shortcuts:
  • ⌘N: New assessment
  • ⌘R: Run audit
  • ⇧⌘R: Re-check
  • ⇧⌘C: Copy Action Items Prompt
  • ⌘?: Help
  • ⌘,: Settings
• More detail is in the other Help sections, or at https://quality.notedsuite.app/help

App Overview

Quality.Noted assesses whether a Mac app is ready to release. It audits the app's source code with Claude Code, scores its readiness out of 100, lists what needs fixing as action items, and builds a walkthrough for what only a person can test.

How an Assessment Works

1. Set up an assessment: where the app will be published, which tool built it, and its project folder.
2. Run the audit. Claude Code checks the app in phases: code, behaviour, then release readiness.
3. Read the results: a readiness score, a verdict, and action items, worst first.
4. Copy the action items into the AI tool that built the app, and fix them.
5. Click Re-check Action Items to confirm the fixes, then work through the Tester Walkthrough.

Who It's For

Anyone shipping a Mac app — especially one built with an AI coding tool. Audits run through Claude Code on your Mac. If the app was built with another AI tool or by hand, Quality.Noted can instead give you prompt files to paste into any AI assistant.

Where Things Are Kept

Each audit is saved in the project's QualityNoted-QA/Results folder, named with its date and time; re-checks go in QualityNoted-QA/Rechecks. Quality.Noted itself only contacts Polar, to activate and check your license, and GitHub, to check for updates; Claude Code does the checking and sends what it reads to Anthropic under your own Claude account. About Quality.Noted ▸ Your Code and Privacy lists everything it keeps.

Setting Up an App

Each app you track starts with a short setup. Your answers shape what Claude looks for.

The Setup Steps

Publishing

Mac App Store, Homebrew or another third-party store, direct download, or still deciding. App Store apps get App Review checks. Homebrew apps get Homebrew's listing rules and end with a verdict: ready for the official list, start with your own tap, or not ready, plus links to Homebrew's docs. Phase 3 is named for your choice.

Tool

Claude Code, another AI tool, or human-coded. Only Claude Code runs checks automatically; for the others you run the audit in your own AI tool (see Using Another AI Tool).

Learning

Let Claude work out the app's features from its code, or describe them yourself.

Folder

The app's project folder — the one containing its source code.

Built artifact

Optional: a built .app or .dmg, so the release check can verify signing, notarization and Gatekeeper directly. A suggestion: leave it empty during development, and use it for a final audit of the build you'll ship, with the release-build switch on.

Questions

A few things only you know: macOS versions you support, known problems, your bar for "ready to ship", and so on.

Review

Confirm everything, pick depth, model and effort, and start.

Using Another AI Tool

Codex works through the prompt files below, like any other AI tool.

1. On the questions step, click Generate a prompt to ask your AI assistant. Paste its whole answer into the box below the button and the fields fill in by themselves.
2. After setup, click Copy Audit Prompt and paste it into your AI tool, with the app's project open if the tool has one. Reply "continue" after Phase 1 and Phase 2.
3. Tools that can save files (Codex, Claude Code, Cursor) send each phase back by themselves. For chat-only tools, reply "continue" once more after Phase 3, copy the results block, and click Paste Results.
4. Click Copy Action Items Prompt and paste it into the same AI chat. Then click Start a new run and audit again in a fresh chat.

Checking the Same App Again

Every run is kept in the sidebar under the app's name. New runs of an app start with the depth, model and effort you used last time for that app.

On the questions step, Or load a previous app fills in the answers you gave last time, so you don't have to retype them.

Files Quality.Noted Adds to Your Project

• CLAUDE.md, in your project folder — the instructions Claude follows, added while a run is going and removed afterwards.
• A QualityNoted-QA folder, holding every run's results, your findings history and your walkthrough answers.

Depth, Model & Effort

Three settings decide how deep a run goes. Depth decides what gets checked. Effort decides how carefully Claude thinks about each thing it checks.

Depth

Quick

Critical issues only — a fast check before you submit. Three phases. Tick Also build a tester walkthrough to add one; it can only be chosen before the run.

Standard

Key flows and high-severity findings. Adds the manual checklist and walkthrough. Five phases.

Thorough

Every source file is read, plus edge cases and accessibility. Five phases. Use this when you need to know the app is ready.

Time depends on the size of your app and the model. A Thorough run on a large app with Opus can take close to two hours.

Model

The main list matches the Claude app's own model menu: Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 4.5. Older models are under More models, for comparing with an earlier run. Default uses whatever Claude Code is set to.

A model card that says "needs Claude Code 2.1.280+" only works once your Claude Code is at least that version.

Effort

Automatic

Follows the depth: Quick uses Medium, Standard uses High, Thorough uses Extra High.

Low

Fastest and cheapest; lightest reasoning.

Medium

Balanced for routine checks.

High

Careful reasoning — a good default for most runs.

Extra High

Recommended for thorough code checks.

Max

Most thorough and most expensive. Use when correctness matters most.

Depth and effort are independent: Thorough with Low still reads every file, just less carefully. Higher effort uses more of your Claude usage.

Running a QA Audit

A run happens in phases. Each phase is a separate Claude session.

Before Phase 1: Build & Tests

On Standard and Thorough, Quality.Noted first builds your app from scratch with your own Swift or Xcode tools, and counts every compiler error and warning. For Swift packages it also runs your app's own tests. This uses no Claude usage, and Claude is given the results as hard facts. It builds into its own cache folder, so it never touches your own builds. Tests for Xcode projects aren't run yet, because they can open your app or control its interface.

The Phases

1. Static Audit — reads the code for bugs, crashes, security and App Store risks.
2. Functional QA — checks how features behave: flows, edge cases, accessibility, error handling.
3. Release Cert — checks readiness for your publishing choice, and sets the readiness score.
4. Manual Checklist — gathers everything a person needs to test by hand. (Standard and Thorough.)
5. Walkthrough — turns that into the step-by-step Tester Walkthrough. (Standard and Thorough.)

Pausing and Continuing

Click Pause Audit at any time, and Continue to pick up later.

Finished phases are always kept. A phase that's interrupted — by Pause or by running out of Claude usage — starts over from its beginning when you continue. If you can, pause just after a phase finishes.

One Run at a Time

Quality.Noted runs one audit or re-check at a time, across all your apps. While one is running, the Run button is disabled. Assessment ▸ Pause Audit works from any screen; the running audit also shows under In progress in the sidebar.

Start another audit while one runs and it goes into the queue: Add to Queue on the last setup step or on an assessment's page. Queued audits show under In queue in the sidebar, in run order. Click ▶ to start one when nothing is running, or right-click to move it up or down or remove it. By default the next one waits for you; Settings ▸ Audits ▸ When an audit finishes can start it automatically after an audit finishes normally.

Running Out of Claude Usage

If a run stops because your Claude usage ran out, Quality.Noted can wait and continue on its own. Turn on Automatically continue paused audits in Settings.

• Out of usage: it tries again every 15 minutes, for up to 12 hours. While you're still out, each try stops immediately.
• Other interruptions, like a network hiccup: it tries again every 5 minutes, up to 3 times.
• Keep Quality.Noted open. You can switch to other screens or apps in the sidebar; the countdown keeps going. If your Mac sleeps, it tries again soon after waking.
• It won't retry a run you paused yourself, or when Claude Code isn't signed in.

Claude Checks Its Own Work

To prove a problem is real, Claude may write and run a tiny test program on your Mac. These run in the background, so you shouldn't see extra Dock icons or "quit unexpectedly" alerts. If you do, it's one of these tests, not your app — click Ignore.

Reading Your Results

When a run finishes, a summary opens with your readiness score and the number of critical and high-severity problems. The full results are in the run's tabs: Overview, Findings (each phase's report, Action Items, and the Prompts Used for reference) and Tester Walkthrough.

The Overview Dashboard

Readiness

A score out of 100, measured against your own bar for "ready to ship".

Critical / High

Serious problems still open. Critical usually blocks a release.

Files Read

How many of your app's source files the run actually examined, e.g. 101/101.

Tokens

How much Claude usage the run used.

Below the cards: which files weren't read and why, your open findings across runs, token usage by run, and your readiness score over time.

Build & Tests

On Standard and Thorough runs, the Build & Tests card shows whether your app built, how many compiler errors and warnings it has (with file and line), and how many of its tests passed. These come straight from your own tools, not from Claude.

Why "Files Read" Matters

"No critical problems" means something very different after reading every file than after reading a few. Quality.Noted always tells you how many were read. If some were skipped, the headline says so and the Coverage card lists them.

How Sure Is Each Finding?

Verified

Claude ran a small test that proved the problem. The proof is saved in the run's Evidence folders.

Proven

The source code alone settles it.

Suspected

Likely, but not proven. Worth checking before acting on it.

Findings Across Runs

Quality.Noted keeps a history of every problem it has found, listed under Open Findings Across Runs on the Overview. Each one gets a stable ID, so the same problem is never counted twice. Click Show all to see every open one, and the copy button beside any of them to get a prompt for fixing just that issue. Every later run must mark each open problem as:

Fixed

The problem is gone.

Still present

The problem is still there.

Not re-checked

The run didn't look — it's still open, not fixed.

Process Notes

Accept Risk: when a finding is true but you won't fix it now, click Accept Risk… on its row: on the re-check scorecard, in the Simple list (Medium and Low sit behind their Show links) or under Open Findings Across Runs. Pick a reason (Deferred to after release, Intended by design or Won't fix), add a note if you like, and confirm. An accepted finding leaves Action Items, the Critical and High counts and re-checks, and later audits are asked not to raise it again. Accepted Risks on the Overview lists them, with Reopen. Low findings need no action: they're never in the fix prompts.

Some findings are about your setup rather than your code — for example, a build that's older than your latest changes. These are reported as process notes and never counted as Critical or High.

Re-check Action Items

After fixing some problems, you don't need another full run to see whether the fixes worked. Re-check Action Items does a much faster, targeted check.

What It Checks

• Every open Critical, High and Medium problem.
• Every open Low in a file you've changed since the last check.
• Those changed files, for new problems your fixes might have caused.

How to Use It

1. Open the app's latest run and click Re-check Action Items. It appears whenever there are open problems.
2. Read the preview: how many problems and changed files it will check, and which model and effort it uses. Nothing runs until you click Re-check.
3. When it finishes, the scorecard shows New problems, Still there, Fixed and Couldn't re-check.
4. Click Copy Remaining Action Items Prompt to copy the Critical, High and Medium problems still left, ready to paste into Claude. Low findings are left out: accept them or keep them for later.

Good to Know

New problems are listed first. A fix that breaks something else is easy to miss.

A re-check updates your findings history, but it doesn't change the Files Read number — only a full run reads every file.

It uses the model and effort of the run you're viewing. If you stop it partway, nothing in your findings history changes.

Tester Walkthrough

Some things only a person can check — how the app looks, feels and behaves. Standard and Thorough runs build a Tester Walkthrough: a checklist of steps to try in the running app.

Testing

1. Open the run's Tester Walkthrough tab.
2. Pick a section at the top. Each item shows where to go, the steps to try, and what should happen.
3. Mark each item Pass, Fail or Needs Improvement, and add a note if something's off.

Your answers save as you go. You can quit and come back any time.

Section Dots

Red

Something in the section failed.

Yellow

Something needs improvement.

Green

Everything in the section is done.

Blue

Partly done.

In Its Own Window

Click Open in Window to pop the walkthrough out, so you can keep it beside the app you're testing.

Who Tests What

The card at the top of the walkthrough shows how many checks are left, how many Claude Code could likely test for you, and how long the rest would take by hand. Walkthroughs from newer runs are already labelled. For older ones, click Estimate: one short Claude call, usually under a dollar.

Claude Tests What It Can…

Saves a brief with only the checks Claude can do and copies a prompt. Paste it into Claude Code; you do the rest here.

Generate Test Files…

One Claude call makes the sample files the checks need, such as a large, corrupt or oddly named file, up to 500 MB in the run's TestFiles folder. Each check lists its files, and Claude's testing brief uses the same ones.

Hand All to Developer…

Saves a brief with every open check and copies a prompt for your developer's Claude Code.

I'll Do It Myself

Shows only the checks that are left.

Claude tests in a Claude Code session you start, and uses your screen, mouse and keyboard while it works. It never passes a check from code alone, marks anything that needs you as Skip with a reason, and won't change your own settings or data. Its results appear here when you switch back to Quality.Noted, and your own clicks never overwrite them.

Each check shows a small tag: Claude can test, Partly Claude, or Needs you. The estimate is a guide, not a promise.

How Big Is It?

Thorough runs build a walkthrough that covers every user-facing feature, so for a large app it can have well over a hundred items. That's intentional — work through it over several sittings.

Sharing & Exports

Everything Quality.Noted finds can be handed to a developer or to Claude.

Prompts to Paste into Claude

Action items

Copy Action Items Prompt copies every Critical, High and Medium action item from the whole run, worst first, ready to paste into your AI tool. It's on the results, the Summary and the Simple view. Each item also has its own copy button.

Walkthrough

Copy Walkthrough Prompt, on the Tester Walkthrough tab, covers only the items you marked Fail or Needs Improvement, with your notes.

Walkthrough briefs

Claude Tests What It Can… and Hand All to Developer…, on the Tester Walkthrough tab, hand open checks to Claude Code to test and fill in.

Re-check

Copy Remaining Action Items Prompt, on the re-check scorecard, covers the Critical, High and Medium problems still left.

One finding

The copy button next to each open finding on the Overview copies a prompt to fix just that one issue — the easiest way to fix things one at a time.

Walkthrough Files

Markdown

A plain-text file.

HTML

Opens in Pages, Word or Google Docs.

Print / PDF

Print it, or save it as a PDF.

Finding the Files

Show in Finder opens the run's folder; Copy Path copies its location. Every audit is saved in your project's QualityNoted-QA/Results folder, named with its date and time — for example Results_2026-10-12_2125. Re-checks go in QualityNoted-QA/Rechecks, as Recheck_2026-10-12_2125. Audits from before October 12, 2026 keep their original names.

Settings

Open Settings with ⌘, or from the Quality.Noted menu. Pick a page on the left.

General

Theme

Light, dark, or follow your Mac.

Text Size

Makes the text you read larger or smaller — results, findings, this Help window and the Tester Walkthrough. A few small labels and badges keep their standard size.

Show results as

Simple or Detailed results.

Audits

When an audit finishes

Whether the next queued audit waits for you or starts by itself.

Auto-continue

Automatically continue paused audits. See Running a QA Audit for exactly when it retries.

Notifications

A macOS notification when an audit or re-check finishes.

Claude Code location

Only needed if claude is installed but wasn't found.

Apps

Rename…

Renames the app in every one of its assessments.

Hide from Sidebar

Out of the sidebar only; it still counts in the Insights Dashboard and Test plans. Show hidden apps from the bottom of the sidebar's Apps list.

Archive

Out of the sidebar, the Insights Dashboard and Test plans. Nothing is deleted; Unarchive brings it back.

Delete…

Removes the app's assessments from Quality.Noted. Its results in the project's QualityNoted-QA folder are kept unless you tick the box to move them to the Trash.

The same actions are on the right-click menu of an app's name in the sidebar.

Transfer

Export All Data… saves every app you track, with its settings and results, into one .zip file. On the other Mac, use Import Data….

Diagnostics

Log activity

Keeps a local log of every Claude Code call — useful if a run fails or gets stuck. Reveal Log in Finder shows it; Clear Log empties it.

Keyboard Shortcuts

Assessments

New Assessment ⌘N

Starts the setup for an app.

Run Audit ⌘R

Runs the audit on the assessment you're viewing. The trust notice still appears first for a new project folder.

Re-check Action Items ⇧⌘R

Re-checks open findings after you've fixed some. Shows a preview first; nothing runs until you confirm.

Copy Action Items Prompt ⇧⌘C

Copies every Critical, High and Medium action item from the run you're viewing.

Pause Audit

Assessment menu. Pauses the running audit from any screen; Continue picks it up later.

Results

Simple Results ⌘1

Score, verdict and action items only.

Detailed Results ⌘2

Every card, chart and finding.

Tester Walkthrough

Pass / Fail / Needs Improvement 1 · 2 · 3

Click a check's title row (or Tab to it), then press 1, 2 or 3. Press the same key again to clear it. Typing in Notes is never affected.

Next Section ⌘]

Moves to the walkthrough's next section.

General

Help ⌘?

Opens this window. The Help menu shows it as ⇧⌘/: the same keys.

Settings ⌘,

Theme, text size, audits, results, logging and data transfer.

Close Window ⌘W

Closes the front window.

Quit ⌘Q

Asks first while an audit is running.

Menu items whose button isn't on screen — Run Audit while an audit is already running, for example — are greyed out.

Printable quick reference →

Troubleshooting

"Not signed in to Claude Code"

Click Sign In to Claude Code, finish signing in, then start the run again.

"The claude CLI wasn't found"

Install Claude Code on this Mac, then reopen Quality.Noted. If it's installed and still not found — for example through nvm or another version manager — run which claude in Terminal, then choose that file in Settings ▸ Audits ▸ Claude Code location.

"Does not support this model"

Your Claude Code is older than the model you picked. Update Claude Code, then try again — or choose a different model. This stops immediately.

"Another run is already in progress"

Only one audit or re-check runs at a time. Wait for it to finish, or pause it. In the sidebar, the running assessment shows a spinner in place of its score; click it to go there.

A Run Stopped Partway

Usually this means your Claude usage ran out. Click Continue once it resets, or turn on auto-continue in Settings. Finished phases are kept; the interrupted phase starts over.

A Run Seems Stuck

On Standard and Thorough, the run first builds your app and runs its tests, which can take up to half an hour on a large project; the progress card says so while it happens. Pausing during that step starts it over on Continue.

Large apps can take a long time on Thorough. If nothing has changed for a long while, click Pause Audit, then Continue. To see what's happening, turn on Log subprocess activity in Settings and check the log.

"Quit Unexpectedly" Alerts During a Run

These come from the small test programs Claude runs to prove a problem, not from your app or Quality.Noted. Click Ignore. Current versions keep these tests hidden.

Glossary of Terms

The words Quality.Noted uses, and what each one means.

Assessments and Audits

ISTQB

International Software Testing Qualifications Board, which defines widely used testing standards, including the seven principles of testing the audit follows. See the Quick Start Guide's Testing approach.

QA

Quality Assurance: checking that software works as intended and meets a release bar before it ships. A Quality.Noted audit is a QA pass done by Claude Code.

Assessment

One app's setup: where it will be published, which tool built it, its project folder, and the depth, model and effort to audit it with. An assessment can be audited many times.

Audit

One full check of the app, run by Claude Code in phases. Each audit is saved as its own dated folder.

Phase

One stage of an audit, run as its own Claude session: Static Audit, Functional QA, Release Cert, Manual Checklist, Walkthrough.

Depth

How thorough an audit is: Quick, Standard or Thorough.

Distribution channel

Where the app will be published — Mac App Store, Homebrew, direct download, or still deciding. It decides what Release Cert checks.

Queue

Audits waiting while another runs. Quality.Noted runs one audit at a time.

Re-check

A fast, targeted audit of open findings and changed files, to confirm fixes.

Results

Readiness score

0–100: how close the app is to your bar for release.

Verdict

Ready to ship, Ready after a few fixes, or Not ready — from the score and open Critical and High findings.

Finding

One problem the audit reports, with a stable ID so it's never counted twice.

Action item

A finding written up as something to fix, ready to hand to your AI tool.

Severity

How bad a finding is if real: Critical, High, Medium or Low. Critical usually blocks a release.

Verification

How sure the audit is: Verified (a test proved it), Proven (the source settles it) or Suspected (likely, not proven).

Process note

An observation about your setup, not your code. Never counted as Critical or High.

Coverage

How many of the app's source files the audit read — shown as Files Read.

Build & Tests

Your app built with your own tools, with its compiler errors, warnings and test results. Exact, and free of Claude usage.

Across Audits

Findings history

Every finding ever reported for a project. Each later audit must account for every open one.

Fixed

The finding is gone.

Still present

The finding is still there.

Not re-checked

The audit didn't look. Still open, not fixed.

Testing by Hand

Tester Walkthrough

A step-by-step checklist for what only a person can check in the running app.

Pass / Fail / Needs Improvement

How a walkthrough check went. Skip marks one that couldn't be tested, with a reason.

Features Available

Auditing

• Three depths: Quick, Standard and Thorough, with your choice of Claude model and effort.
• Build & Tests before Phase 1, using your own tools — exact results.
• Release checks for the Mac App Store, Homebrew, or direct download, with guidance for regulated content.
• A queue for audits started while another runs, and automatic continuing after Claude usage runs out.

Results

• Readiness score and verdict, in a Simple or Detailed view.
• Every finding labelled by severity and by how sure the audit is.
• Files Read, so you know what a clean result covered.
• A findings history across audits, so nothing quietly drops off the list.

Fixing

• Copy Action Items Prompt: one prompt with everything to fix, worst first, for the AI tool that built the app.
• A copy button on each finding, to fix one at a time.
• Re-check Action Items: confirms fixes in minutes instead of a full audit.

Testing by Hand

• Tester Walkthrough with Pass, Fail and Needs Improvement, saved as you go.
• Hand checks to Claude Code to test for you, or to your developer.
• Export as Markdown, HTML or PDF.

Other AI Tools

• Prompt files for ChatGPT Codex, Gemini Code Assist, Cursor, other AI tools, or human-coded apps.

Your Data

• Results are saved in your project folder; your assessments in ~/Library/Application Support/QualityNoted. Export All Data… moves everything to another Mac.

What's New

New in Quality.Noted 27.1.

Highlights

A QA audit run by Claude Code

Quality.Noted drives the claude CLI headlessly in your project, one phase at a time, on your own Claude plan.

Readiness score

Scored out of 100 against your distribution channel: App Store, Homebrew or direct download. 85 or more is ready; read Critical and High first.

Action items, then re-check

Copy Action Items Prompt hands every finding to Claude Code to fix. Re-check Action Items confirms the fixes for a fraction of a full run.

Tester Walkthrough

A manual test plan for what only a person can verify, with the checks Claude can do handed to Claude Code.

Insights Dashboard

Readiness over time and fix rate, across one app or several.

Release Notes

Version 27.1

The first release of Quality.Noted. The last number in the version is the build.

• Audits in phases with Claude Code, with Build & Tests, readiness score and verdict.
• Simple and Detailed results; findings history across audits.
• Action items, Re-check Action Items, and the Tester Walkthrough.
• Release checks for the Mac App Store, Homebrew and direct download.
• Audit queue, automatic continuing, keyboard shortcuts.