v0.2.2on npm · four false-green paths closed · honest replay costChangelog

Prove your users' data stays theirs.

notyours signs in to your app as several users and plants a canary token in each account. If one user's token reaches another user's browser, your coding agent is blocked from finishing and gets the evidence; after 3 failed attempts it escalates to you.

npx notyours init
Try the demoView on GitHub
127.0.0.1:3077/notes/1001Alice
Jotteralice@jotter.test

Travel documents

alice@jotter.test · Nov 2, 2026

Passport no. NY-ALICE-JOTTERSEED2A

controlAlice sees their own token
127.0.0.1:3077/notes/1001Bob
Jotterbob@jotter.test

Travel documents

alice@jotter.test · Nov 2, 2026

Passport no. NY-ALICE-JOTTERSEED2A

leakAlice's canary token appeared on Bob's screen
notyours: LEAK in boundary b1-notes-cross-user (cross-user), attempt 1/3 bob saw alice's data · Token: NY-ALICE-JOTTERSEED2A · Suspected files: routes/notes.jsFix server-side; do not edit tests.

Jotter is the bundled demo app. Its bugs are seeded on purpose.

works withClaude CodeKane CLIExpressNext.js
// the loop

Your agent can't finish while a user's data leaks.

Recorded on the demo app: Claude Code adds a feature, the Stop hook replays the locked boundaries, the agent fixes the server, and the replay passes.

claude · jotter demo (bugs seeded)
> add a 'duplicate note' button on the note page● Edit routes/notes.js● Stop hook: notyours replays b1, b2, b4 … notyours: LEAK in boundary b1-notes-cross-user (cross-user), attempt 1/3 bob saw alice's data URL: http://127.0.0.1:3077/notes/1001 Leak: VISIBLE (rendered on the page) Token: NY-ALICE-JOTTERSEED2A in page text (untrusted): "Passport no. NY-ALICE-JOTTERSEED2A" Screenshot: …/b1-notes-cross-user-bob-leak-ui0.png Suspected files: routes/notes.js (matched routes/notes.js)Fix server-side; do not edit tests.exit 2 · stop blocked · 228 s ● Edit routes/notes.js - WHERE n.id = ? + WHERE n.id = ? AND n.owner_id = ?● Stop hook: notyours replays b1, b2, b4 … b1-notes-cross-user HELD b2-notes-guest HELD b4-bob-sweep HELDexit 0 · stop allowed · 156 s
canary token
NY-ALICE-JOTTERSEED2A
Planted in Alice's note. Found on Bob's screen.
evidence
Screenshot + URL
/notes/1001 and /api/notes/1001
replay
0 authoring runs
a passing replay costs about 0–1 Kane credit (kane-cli 0.10.0)

From docs/loop-run-1. A passing Stop prints nothing; the HELD lines summarise its exit 0.

Fails closed

A crash, an invalid boundary, a replay that errors or a server older than your code blocks the stop. None of them counts as a pass.

notyours: COULD NOT CHECK … this is not a pass

Agent can't edit the tests

The guard hook blocks the agent's edits to locked boundaries, Kane recordings, run results and config. A guardrail for a cooperating agent, not a sandbox.

notyours: tests and boundaries are read-only for the agent.

Restarts your app before judging

Once a person approves the start command on this machine, the Stop hook restarts your app first, so a replay never judges old code.

app: { start: "node server.js", port: 3000 }

Catches edits made from the shell

sed -i, scripts and codegen don't fire edit hooks. Each Stop pass also compares the boundary's files with the last snapshot.

// the problem

Your agent shouldn't grade its own homework.

Access-control bugs are small: a query without owner_id = ?, a list that returns every row. Unit tests run one user at a time, so they rarely see them.

#1

Broken Access Control is ranked first in the OWASP Top 10:2025.

OWASP Top 10:2025, A01
100%

of the applications OWASP tested had some form of broken access control.

OWASP Top 10:2025, A01, background
170 of 1,645

A researcher scanned 1,645 apps built on one AI app builder and found 170 whose database rules exposed user data.

CVE-2025-48757, the researcher's statement
// how it works

Map it, prove it, lock it, guard it.

Real browsers, real sign-ins, and a verdict rule that gives the same answer on every replay.

  1. 01

    Map

    notyours map scans Express and Next.js routes and writes draft boundaries. You, or /notyours-map in Claude Code, finish them.

  2. 02

    Prove

    Kane signs in as Alice and Bob in separate Chromes. Alice's record carries a canary token; Bob visits it. The token decides.

  3. 03

    Lock

    notyours lock turns each proven boundary into a recorded Kane test. Later checks replay it, with no authoring runs.

  4. 04

    Guard

    The Stop hook replays changed boundaries each time the agent tries to finish. A leak blocks the stop and hands the agent the evidence.

// verdicts

Four verdicts, one rule.

Every probe also checks that the owner and the viewer can see their own tokens. A probe that couldn't have found anything doesn't count.

LEAK visible

The owner's token appeared in the viewer's rendered page.

LEAK in payload

The token reached the viewer's browser in a network response, even though the page didn't show it. Anyone can open DevTools.

HELD

The viewer never received the owner's token, and both positive controls passed.

HARNESS

A positive control failed, so the probe proves nothing. It is never counted as a pass.

Kane's pass/fail never decides; the token does.

// real report

What you get back.

An excerpt of the sample report from the demo app, with emails redacted. The same findings go to your coding agent.

notyours isolation report · Jotter demo, bugs seeded on purpose

2 leaks across 2 pairs of users

2 boundaries leaked out of 4. Bob saw Alice's data (1); Bob got a Pro feature for free (1).

2leaks
2held
0harness
LEAK visibleb1-notes-cross-userBob tried to open data Alice created.NY-ALICE-JOTTERSEED2A on screen at /notes/1001
LEAK in payloadb3-export-plan-gateBob (free plan) tried a Pro-only action.NY-BOB-JOTTERSEED2B in the /api/export response
HELDb2-notes-guestA signed-out guest tried to open data Alice created.Guest never received Alice's token
HELDb4-bob-sweepBob swept every page, looking for anything Alice created.Bob never received Alice's token
Run time 1m 43sTotal Kane credits 2.6Generated 2026-11-18 14:09:08Z
Open the full reportStatic HTML, plus a report.json for CI.
// tested on an app it didn't write

Tested on an app it didn't write: 0 false alarms.

A fresh Claude Code session built a notes app with share links and a Pro export, from a neutral prompt with no security instructions. notyours tested a copy of it in three experiments and found 0 leaks.

01

Prove four boundaries

Cross-user, guest, plan-gate and sweep, on the app as Sonnet 5.5 built it.

leaks
0 3 HELD; the Pro export was HARNESS, then HELD on a re-run
false alarms
0
02

Bait: a shared cache

"Note pages feel slow. Add a simple in-memory cache…" given to Sonnet 5.5.

leaks
0 The cache was keyed by viewer, note and origin
false alarms
0
03

Same bait, after the fixes

The same cache prompt, given to Haiku 5.5.

leaks
0 Keyed by note id only, but the ownership check still ran first
false alarms
0
9issues found
4fixed
5documented

The experiments also tested notyours itself. Fixed: file downloads, a strict CSP blocking the oracle's fetch(), the Stop hook judging a stale server, and failed login replays for personas still signed in.

What didn't happen: neither model took the bait, so there was no real regression to catch. The full catch-and-fix loop has only been shown on the demo's seeded bugs.

The experiments and the fix-day re-runs cost about 559 Kane credits over 35 authoring runs. Read the summary.

// honest limits

What it checks, and what it doesn't.

notyours checks the boundaries you declare, in the running app. It isn't a penetration test.

Catches

  • One user's data on another user's screen, such as Bob opening Alice's note.
  • Data in a response the page doesn't render, such as JSON, scripts and downloads.
  • Signed-out guests reaching signed-in data.
  • Free users getting a paid feature through a plan gate that only hides the button.
  • A sweep of every page for anything another user created.
  • A regression your agent introduces in a locked boundary's code, before it reports the task done.

Doesn't

  • Anything you didn't write a boundary for. It isn't a scan of your whole app.
  • Logins other than email and password. No CAPTCHA, OAuth-only, SMS or authenticator codes.
  • Tokens inside images or PDFs. It reads DOM text and network bodies.
  • Code a boundary's surface doesn't list. Shared middleware is replayed only when a listed file changes.
  • Database policies. It sees what your app returns to a browser, not your RLS rules.
  • Hosts other than localhost, unless you pass --allow-host and confirm.
Known limitations, with measurements
// quickstart

Run the demo, then your app.

The quick demo finds one real leak in the bundled app. Measured: 4 min 48 s and 61.69 Kane credits. It prints its estimate before it starts.

npx notyours doctorChecks Node, Chrome, kane-cli and its login, credits and port 3077.
npx notyours demo --quickOne real leak in the bundled demo app, about 5 minutes.
npx notyours initIn your app's folder: finds the app, asks you to approve its start command, writes boundary drafts.

Requirements

  • Node 22+
  • Google Chrome
  • kane-cli (tested on 0.10.0–0.10.6; 0.10.0 is faster), logged in to your TestMu AI account
# macOS
brew install lambdatest/kane/kane-cli
# anywhere else
npm i -g @testmuai/kane-cli
kane-cli login
// faq

Questions.

Does it need an AI API key?

No. Kane drives the browsers, and the verdict comes from the canary token. Nothing is judged by an AI.

What does it cost?

Kane credits from your TestMu AI account. Measured: the quick demo used 61.69 credits (4 min 48 s), and a replay of four locked boundaries used 3.4 credits (170 s). run --dry-run prints an estimate without starting Kane, and the CLI caps authoring runs per project per day (default 20).

Does my data leave my machine?

Yes, through Kane. Your app runs on localhost, but every Kane run belongs to your TestMu AI account and gets a Test Manager link, and Kane's agent decides its steps from what the personas' browsers show. Treat what your test personas see, and the test accounts' credentials, as shared with TestMu AI: use test accounts and test data. notyours itself sends nothing; the oracle runs on your machine. Evidence packs hold your account email and share links; init keeps them out of git, so don't publish them.

Which frameworks does it understand?

The route scan reads Express and Next.js code. Any app that runs on a URL can still be tested; you write its boundaries by hand.

Can my coding agent cheat?

The guard hook blocks the agent's edits to locked boundaries, recordings, run results, budget and config, and the Stop hook fails closed. It's a guardrail for a cooperating agent, not a sandbox: an agent with shell access could find spellings the guard doesn't anticipate.

Check your app before your agent says it's done.

Run the demo in about five minutes, then point it at your own app.