Verilay Report
Analyse your own app →
Deep scan report · GitHubUp to 150 files read, every risky package named
List this app in Checked apps →
juice-shop/juice-shop
Scanned 31 Oct 2026, 14:12 UTC
OWASP Juice Shop is an intentionally insecure web shop application used for security training, CTFs and awareness demos
Not Ready
Several critical findings — fix these before launching with real users.
IT Review Pack → a printable checklist for IT teams approving this app
F
Score
12
Critical
27
Warnings
6 of 6
Areas checked
← Back to the summary
Fix prompts
Copy a prompt, paste it into your AI builder (Lovable, Replit, Bolt, Cursor…), then re-run Verilay to check it worked. Treat each one as something to verify, not an order: your AI builder knows your code.
1. Private signing key committed in source codecriticalAuth
A private key embedded in lib/insecurity.ts (line 21), the file that signs login tokens, lets anyone who can read the repository forge valid login passes for any user, including admins.
How: 1. Open lib/insecurity.ts and confirm what the key at line 21 is used for. 2. Search the whole repository and its git history for other private keys or key-like strings. 3. Decide how the key should be supplied (platform secret store) and whether the old key must be treated as compromised and rotated.
Effort: 1 hour
Step 1: paste this into your AI builder. It explains the problem first and changes nothing.
I received a security review flagging a private signing key committed in lib/insecurity.ts at line 21, which is used to sign login tokens. Please review and advise without making changes yet. Explain how the key is used for signing and verifying tokens. Then check the ENTIRE repository (including config files, scripts and git history if you can see it) for any other embedded private keys, PEM blocks, or key-like strings, and LIST each file and line where you find one. Advise how to move the key into the platform's managed secret storage without breaking existing code, and whether the current key should be rotated and treated as compromised. Suggest the smallest safe change only if genuinely needed. Please don't change anything yet. In plain English, tell me: 1. Is this really a problem? Show me the exact file and lines. 2. Who could misuse it, and what could they see or do? 3. How is this part meant to be used? What relies on it working the way it does now: other pages, a mobile app that is already released, scheduled jobs, webhooks, emails, payments or sign-in? 4. What could the fix break, and how would we avoid that? 5. How will we check afterwards that everything still works, on the website and in the mobile app if there is one? Also check the whole app for the same kind of problem, and list every place with the file and line. Don't change anything until I reply.
Step 2: only when its answer makes sense to you, paste this to let it make the change.
Thanks. Please fix only the place named in this report, not the other places you listed: we'll do those separately, one at a time. Keep everything you listed working. Don't publish or deploy anything. I'll test it in the preview first. Then tell me exactly what you changed, file by file, and walk me through the checks from your answer so I can confirm everything still works. If there's a mobile app that's already released, remind me to test it after publishing.
Check it yourself · about 3 minutes
Is a secret saved in your GitHub repository?
  1. Open your repository on GitHub.
  2. Open the file and line that Verilay named in this report.
  3. Decide whether the value is real, or an example or placeholder.
✓ It's an example, a placeholder, or a value that is meant to be public.
⚠ It's a real key. Make a new key in the service's dashboard and switch to it. Deleting the file isn't enough, because the old key stays in the repository's history.
Never paste a real key into a chat, an email or a website while checking.
Only try this on your own app, with test accounts. Never on someone else's app.
2. Open registration as admin and missing authorization checkscriticalAuth · Express
POST /api/Users accepts a role field from the request (admin works), PUT /api/Products/:id has its isAuthorized() guard commented out, and the password-reset rate limiter trusts the client-controlled X-Forwarded-For header. Anyone can become an admin or edit data without logging in.
How: 1. Review the user creation route and see how role is handled. 2. Check the product update route and why the guard is commented out. 3. Review how the rate limiter identifies clients.
Effort: 1 hour
Step 1: paste this into your AI builder. It explains the problem first and changes nothing.
I received a security review flagging three authorization problems: POST /api/Users accepts a role field from the request body (admin works), PUT /api/Products/:id has its isAuthorized() guard commented out, and the password-reset rate limiter keys on the client-controlled X-Forwarded-For header. Please review and advise without making changes yet. For each, explain the risk and the minimal safe change using the auth middleware already in the project. Then, for EVERY route in the repository, produce a table listing the route, HTTP method, whether it requires authentication, whether it checks the caller's role or ownership, and whether it accepts fields like role or isAdmin from the request body. LIST every route that lacks a check or lets clients set privileged fields. Please don't change anything yet. In plain English, tell me: 1. Is this really a problem? Show me the exact file and lines. 2. Who could misuse it, and what could they see or do? 3. How is this part meant to be used? What relies on it working the way it does now: other pages, a mobile app that is already released, scheduled jobs, webhooks, emails, payments or sign-in? 4. What could the fix break, and how would we avoid that? 5. How will we check afterwards that everything still works, on the website and in the mobile app if there is one? Also check the whole app for the same kind of problem, and list every place with the file and line. Don't change anything until I reply.
Step 2: only when its answer makes sense to you, paste this to let it make the change.
Thanks. Please fix only the place named in this report, not the other places you listed: we'll do those separately, one at a time. Keep everything you listed working. Don't publish or deploy anything. I'll test it in the preview first. Then tell me exactly what you changed, file by file, and walk me through the checks from your answer so I can confirm everything still works. If there's a mobile app that's already released, remind me to test it after publishing.
Check it yourself · about 3 minutes
Can someone try passwords without limit?
  1. On your own login page, use a test account.
  2. Enter a wrong password about ten times in a row.
  3. Watch for a pause, a "too many attempts" message, or a block.
✓ You're slowed down or blocked.
⚠ Every try is accepted, however many. Tell your AI builder.
Use a test account, not your real one, in case it locks.
Only try this on your own app, with test accounts. Never on someone else's app.
3. Hard-coded cookie secret and publicly browsable sensitive folderscriticalConfig · Express
server.ts signs cookies with the literal secret 'kekse', allows every origin via cors(), and exposes /ftp, /encryptionkeys, /support/logs and /infrastructure with directory listings, plus a public /metrics endpoint. This leaks keys and logs and makes cookies forgeable.
How: 1. Review server.ts for the cookie secret, CORS setup and static folder exposure. 2. Check what files are reachable under the exposed folders. 3. Decide which routes should be removed or restricted.
Effort: 1 hour
Step 1: paste this into your AI builder. It explains the problem first and changes nothing.
I received a security review flagging server.ts for a hardcoded cookie secret ('kekse'), app.use(cors()) allowing all origins, directory listings on /ftp, /encryptionkeys, /support/logs and /infrastructure, and a public /metrics endpoint. Please review and advise without making changes yet. Explain what each exposure reveals and the minimal change to fix it using the existing Express setup (no new libraries). Then check the ENTIRE repository for other static-serving, directory listing, CORS, or hard-coded secret usages, and LIST each with file name and what it exposes. Please don't change anything yet. In plain English, tell me: 1. Is this really a problem? Show me the exact file and lines. 2. Who could misuse it, and what could they see or do? 3. How is this part meant to be used? What relies on it working the way it does now: other pages, a mobile app that is already released, scheduled jobs, webhooks, emails, payments or sign-in? 4. What could the fix break, and how would we avoid that? 5. How will we check afterwards that everything still works, on the website and in the mobile app if there is one? Also check the whole app for the same kind of problem, and list every place with the file and line. Don't change anything until I reply.
Step 2: only when its answer makes sense to you, paste this to let it make the change.
Thanks. Please fix only the place named in this report, not the other places you listed: we'll do those separately, one at a time. Keep everything you listed working. Don't publish or deploy anything. I'll test it in the preview first. Then tell me exactly what you changed, file by file, and walk me through the checks from your answer so I can confirm everything still works. If there's a mobile app that's already released, remind me to test it after publishing.
Check it yourself · about 6 minutes
Is a secret saved in your GitHub repository?
  1. Open your repository on GitHub.
  2. Open the file and line that Verilay named in this report.
  3. Decide whether the value is real, or an example or placeholder.
✓ It's an example, a placeholder, or a value that is meant to be public.
⚠ It's a real key. Make a new key in the service's dashboard and switch to it. Deleting the file isn't enough, because the old key stays in the repository's history.
Never paste a real key into a chat, an email or a website while checking.
Does the live site send the standard safety headers?
  1. Go to a free checker such as securityheaders.com and type the address of your own live site.
  2. Read the grade and the list of missing headers.
✓ The main headers are present.
⚠ Some are missing. Ask your AI builder or your host how to add them. Many are optional.
These headers are set by your hosting, not in the app's code. Only check your own site.
Only try this on your own app, with test accounts. Never on someone else's app.
4. Private key files committed in Terraform infrastructure filescriticalConfig
The secret scan found private keys at line 171 of infrastructure/terraform/networking.tf and terraform/networking.tf. These may protect servers or network access, and anyone who reads the repository can use them.
How: 1. Open both files at line 171 and identify what the key is for. 2. Determine whether it is a real credential or sample/placeholder content. 3. If real, plan rotation and removal from git history.
Effort:
Step 1: paste this into your AI builder. It explains the problem first and changes nothing.
I received a security review flagging this problem: Private key files committed in Terraform infrastructure files. Details from the review: The secret scan found private keys at line 171 of infrastructure/terraform/networking.tf and terraform/networking.tf. These may protect servers or network access, and anyone who reads the repository can use them. Please review and advise without making changes yet. Please don't change anything yet. In plain English, tell me: 1. Is this really a problem? Show me the exact file and lines. 2. Who could misuse it, and what could they see or do? 3. How is this part meant to be used? What relies on it working the way it does now: other pages, a mobile app that is already released, scheduled jobs, webhooks, emails, payments or sign-in? 4. What could the fix break, and how would we avoid that? 5. How will we check afterwards that everything still works, on the website and in the mobile app if there is one? Also check the whole app for the same kind of problem, and list every place with the file and line. Don't change anything until I reply.
Step 2: only when its answer makes sense to you, paste this to let it make the change.
Thanks. Please fix only the place named in this report, not the other places you listed: we'll do those separately, one at a time. Keep everything you listed working. Don't publish or deploy anything. I'll test it in the preview first. Then tell me exactly what you changed, file by file, and walk me through the checks from your answer so I can confirm everything still works. If there's a mobile app that's already released, remind me to test it after publishing.
Check it yourself · about 3 minutes
Is a secret saved in your GitHub repository?
  1. Open your repository on GitHub.
  2. Open the file and line that Verilay named in this report.
  3. Decide whether the value is real, or an example or placeholder.
✓ It's an example, a placeholder, or a value that is meant to be public.
⚠ It's a real key. Make a new key in the service's dashboard and switch to it. Deleting the file isn't enough, because the old key stays in the repository's history.
Never paste a real key into a chat, an email or a website while checking.
Only try this on your own app, with test accounts. Never on someone else's app.
What we found, by area
Authcritical
💡 Think of a farmers' market stall-permit office: each trader gets a stamped badge saying whether they are a customer, a stall holder or the manager. Here, the office will stamp 'manager' on your badge if you simply ask, one gate has had its badge check removed, and the stamp itself has been left lying on the counter.
Login uses signed digital passes (JWT) plus an optional phone-code second step, but the check on who you are is leaky in several places: signup accepts a self-chosen admin role, login can be tricked with typed text, password change can skip the old-password check, Google sign-in accounts get a guessable password, and signing keys sit in the repository. Note: this is OWASP Juice Shop, a deliberately vulnerable training app, so many of these are intentional, but each would be serious in a real shop.
Private signing key committed in the code (confirmed scan) — The secret scan found a private key file embedded in lib/insecurity.ts at line 21, the file that signs login passes. The same type of private key was also found at line 171 of infrastructure/terraform/networking.tf and terraform/networking.tf (listed under Config). Anyone who can read the repository can use this key to forge login passes.
Login bypass via SQL injection — routes/login.ts pastes the typed email straight into a raw database query, so input like ' OR 1=1-- logs in as the first user (often an admin) with no password.
Anyone can register as an admin and some routes lack login checks — POST /api/Users accepts a role field from the request (admin works). PUT /api/Products/:id has its isAuthorized() guard commented out. The password-reset rate limiter keys on the client-controlled X-Forwarded-For header, so it can be dodged.
Password can be changed without knowing the current one — routes/changePassword.ts only checks the current password if one is supplied, so omitting the 'current' parameter skips the check. New passwords also travel in the web address.
Predictable password for Google sign-in accounts — frontend/src/app/oauth/oauth.component.ts builds each Google user's password as btoa(reverse(email)), so anyone who knows an email can compute the password and log in through the normal login form. Related weaker points: the 2FA secret is stored as a plain string in models/user.ts (warning), reviews accept any author name, and login tokens sit in localStorage and readable cookies.
💡 Key concept: Never trust anything a visitor says about themselves, such as their role or user ID. The server must decide who is who, and every sensitive action must re-check it.
Configcritical
💡 Like the whiteboard behind a juice bar counter with recipe ratios, opening hours and the till code: staff read from it daily, but the till code should never be taped to the shop window, which is what happens here.
Settings come from a config library and environment variables, which is a good habit, but the repository contains private keys in infrastructure files, a hardcoded cookie secret, test passwords in code, open cross-origin access, and publicly browsable folders for keys and logs. The whole config is also served to every visitor.
Private key files committed in infrastructure setup (confirmed scan) — The secret scan found a private key file at line 171 of infrastructure/terraform/networking.tf and again at line 171 of terraform/networking.tf. These are infrastructure setup files, so the keys may protect servers or network access.
Open cross-origin access, hardcoded cookie secret and publicly browsable sensitive folders — server.ts uses app.use(cors()) allowing every origin, signs cookies with the literal secret 'kekse', and exposes /ftp, /encryptionkeys, /support/logs and /infrastructure with directory listings. /metrics is also public.
Hard-coded credentials in source code — routes/login.ts and routes/resetPassword.ts embed account passwords and security answers as plain strings in comparison logic. The login page component (frontend/src/app/login/login.component.ts) also ships testingUsername and testingPassword to every visitor's browser.
Entire app configuration served to any visitor — routes/appConfiguration.ts returns the whole config object to any caller, removing only llmApiUrl. Any sensitive setting added later would leak, and LLM_API_KEY silently falls back to an empty string.
💡 Key concept: Secrets and private folders belong behind a locked door, never in code, in the repository, or on public web addresses.
Databasecritical
💡 Picture a juice bar's order-ticket rail and loyalty-card box: every customer card, basket slip and delivery note lives there. The problem is that a customer can shout a reworded request over the counter and the clerk fetches the whole ledger.
The app stores users, baskets, addresses, cards, reviews and orders in SQLite (via Sequelize) and MongoDB. Product search builds raw SQL from typed text, and review lookups run text as a database mini-program, which can expose all customer data. Auto-generated endpoints expose many tables, and the database is wiped on every start. No secret-scan or dependency issue points directly at the database itself, though the Sequelize library has known SQL injection flaws (see Libraries).
SQL injection in product search and review lookups exposes all data — routes/search.ts concatenates req.query.q into a raw SELECT, allowing UNION queries that return the Users table (emails and password hashes) and the schema via sqlite_master. showProductReviews.ts, trackOrder.ts and routes/chat.ts put user input into MongoDB $where JavaScript expressions, and review creation stores unvalidated input and a client-supplied author.
Auto-generated endpoints expose sensitive tables and data is wiped on start — finale-rest generates create/read/update/delete routes for Users, Cards, Addresses, Feedbacks, Complaints and more, excluding only password and totpSecret. Some routes (/api/Feedbacks, /api/Deliverys, GET /api/Products) are unauthenticated, /api/Quantitys is guarded by an IP filter on an invalid address, and sequelize.sync({ force: true }) drops all data at every start. Address routes also trust a UserId sent in the request body.
Sensitive user fields stored with little protection — The Users model stores password via a custom hash helper, plus totpSecret, lastLoginIp and deluxeToken as plain strings, and Address stores mobile number as an integer with loose validation.
💡 Key concept: The database should only receive questions your code wrote, with visitor input passed in as plain values, never as part of the question.
APIcritical
💡 Like a hotel concierge desk with a key cabinet: guests ask for things, and the concierge decides who gets a key. Here, several cabinet doors were deliberately left unlocked and some desks hand out keys based on what the guest says rather than what the booking shows.
The Express API handles login, search, baskets, orders, reviews, uploads, data export and a chatbot. It contains deliberate flaws: queries built from typed text, routes that trust user-supplied IDs, a data export that leaks other users' memories, an unlocked product-edit route, open file and metrics endpoints, and prices computed in the browser. Several of these overlap with Auth and Database findings (login and search injection are described there) and would be critical in a real deployment.
Missing auth on product updates plus open file and metrics endpoints — In server.ts the protection on PUT /api/Products/:id is commented out, CORS is fully open, and /ftp, /support/logs, /encryptionkeys and /metrics are served with directory listing and no authentication. Auto-generated finale-rest endpoints expose full create/read/update/delete on many models.
Data export reads any user's memories via client-supplied UserId — routes/dataExport.ts identifies the user from the Authorization header but then fetches memories using req.body.UserId, which the caller controls. Orders are also looked up by a vowel-masked email, which is weak and guessable. routes/address.ts similarly filters by req.body.UserId.
Risky handling of uploads, templates and server-side web requests — fileUpload.ts extracts zip entries using their own file names and parses XML and YAML from uploads. dataErasure.ts passes ...req.body into a template render, letting callers influence which server file is used for the page layout (guarded only by a weak keyword block-list). profileImageUrlUpload.ts downloads a user-supplied web address and stores files, including SVG, in a publicly served folder.
Prices and discounts calculated in the browser — frontend/src/app/payment/payment.component.ts computes totalPrice from sessionStorage values and a hard-coded campaigns table, and the wallet top-up sends balance: this.totalPrice from the client. Unless the server recalculates, a user can edit these values.
Outbound webhook sends server details with no signing or timeout — lib/webhook.ts POSTs JSON to the SOLUTIONS_WEBHOOK address including hostname, OS details, app version, a cheat score and the CTF flag. There is no signature, no timeout, no error handling, and the full URL (which may contain a secret) is logged.
💡 Key concept: Every route must decide who is allowed to use it and must take identity from the verified login, never from fields the caller sends.
Frontendwarning
💡 Like the front-of-house floor of a theatre or juice bar: the displays, counters and programmes customers deal with directly, while the vault and records stay out of sight. Customers can scribble on anything they hold, so the back office must never trust it.
The Angular storefront shows products, basket, checkout, login and tutorials. It stores the login token in script-readable storage, ships a test account's credentials in the login page, injects tutorial hints with innerHTML, and fetches secrets for 2FA setup into the browser. The route that serves the built front page is safe. The confirmed secret scan also found sample login tokens in front-end test files (see the first finding).
Sample login tokens (JWTs) in front-end test files (confirmed scan) — The secret scan found JSON Web Tokens at frontend/src/app/app.guard.spec.ts line 46, and frontend/src/app/last-login-ip/last-login-ip.component.spec.ts lines 72 and 78. These are in test files and are most likely sample tokens used to exercise the code, but they should be confirmed as fake and unsigned by a real key.
Login token stored in script-readable storage and test credentials in code — login.component.ts, two-factor-auth-enter.component.ts and payment.component.ts save the token to localStorage and a JavaScript-set cookie (no HttpOnly flag), so any malicious script on the page could read it. login.component.ts also hard-codes testingUsername and testingPassword. The Google OAuth client ID is public by design and fine.
User text and files reach pages without cleaning (XSS risk) — saveLoginIp.ts stores the true-client-ip header unsanitised, trackOrder.ts echoes the id back, and review text is saved as-is. If pages show these without escaping, a script can run in other visitors' browsers. profileImageUrlUpload.ts writes downloaded files, including SVG, into the public frontend assets uploads folder. hacking-instructor/index.ts also uses innerHTML for developer-written hint text, which is fine today but risky if user text ever reaches it.
Two-factor secret handed to the browser and fragile wallet page — two-factor-auth.component.ts fetches the authenticator secret and setup token from the status endpoint and builds the otpauth URL in the browser, logging failures only to console.log. web3-sandbox.component.ts calls window.ethereum.on(...) before checking that it exists, which throws an error for visitors without a wallet extension.
💡 Key concept: Anything in the browser can be seen and changed by the person using it, so it should hold convenience data only, never secrets or final decisions.
Librariescritical
💡 Like a restaurant buying bread from a baker, cheese from a dairy and sauces from a specialist: you rely on outside suppliers, and if one ships spoiled goods your dishes are affected, so you must watch for recalls. Here several suppliers have open recalls.
The dependency check covered 170 directly installed packages (not the packages those depend on, which could not be checked because no full package list was found) and confirmed 3 critical, 17 high, 26 moderate and 1 low advisories. The worst involve very old login and database packages that are pinned deliberately for training, so they would be serious in a real shop. Two dev-only tools are also flagged and are lower-priority housekeeping.
Critical flaws in jsonwebtoken, marsdb and sequelize — [email protected] has a verification bypass (GHSA-c7hr-j4mj-j2w6, fix: upgrade to 4.2.2; later advisories say upgrade to 9.0.0). [email protected] has command injection (GHSA-5mrr-rgp6-x4gr, no fix published yet). [email protected] has SQL injection for Oracle databases (GHSA-v8fg-2rw7-q452, fix: upgrade to 6.37.4) and a high-severity SQL injection via JSON column cast type (GHSA-6457-6jrx-69cr, fix: upgrade to 6.37.8).
High-severity flaws in login, text-cleaning and server packages — [email protected] allows authorization bypass (GHSA-6g6m-m6h5-w9gf, upgrade to 6.0.0). [email protected] allows legacy key use (GHSA-8cf7-32gw-wr33, upgrade to 9.0.0). [email protected] is open to ReDoS slowdowns (GHSA-cgfm-xwp7-2cvr, upgrade to 2.7.1). [email protected] has a denial of service via trailing bytes (GHSA-27r3-jxmh-p828, upgrade to 1.20.9). [email protected] leaks memory (GHSA-vc2v-76pw-4v95, upgrade to 1.8.2). [email protected] has a denial of service (GHSA-q7qp-rjg7-q26m, upgrade to 2.5.0). [email protected] can exhaust memory (GHSA-mrhx-c7xc-m3w5, upgrade to 4.6.6). [email protected] has code injection via _.template (GHSA-r5fr-rjxr-66jc, upgrade to 4.18.0). [email protected] has CLI command injection (GHSA-5j98-mcp5-4vw2, upgrade to 10.5.0). [email protected] has three CPU-exhaustion flaws (GHSA-2883-xcg3-v3hh to 3.15.2, GHSA-52cp-r559-cp3m to 3.15.0, GHSA-5p4m-2wfm-xmqj to 3.15.1).
High-severity flaws in Angular packages — @angular/[email protected] has a cache-key ambiguity that can reuse responses across requests (GHSA-jhpw-976m-542j, upgrade to 22.0.2). @angular/[email protected] has three denial of service flaws in server-side rendering (GHSA-57xq-rjx2-v5xh and GHSA-62vg-58rm-qff7, upgrade to 22.2.1; GHSA-ff3f-86qr-9cv3, upgrade to 22.2.0).
Moderate and low issues across many packages — Moderate: @angular/common (GHSA-p297-fm68-3q8c, to 22.1.1), @angular/compiler and @angular/core sanitization bypass (GHSA-hh8m-fm6v-7cvg, to 22.1.0), file-type infinite loop (GHSA-5v7r-6r5c-r473, to 21.3.1), glob abort bypass (GHSA-3g3m-2fjr-6pw4, to 14.0.2), hbs sensitive information exposure (GHSA-7f5c-rpf4-86p8, no fix yet), js-yaml (GHSA-h67p-54hq-rp68 to 3.15.0, GHSA-mh29-5h37-fv8m to 3.14.2), jsonwebtoken key and algorithm flaws (GHSA-hjrf-2m68-5959, GHSA-qwph-4952-7xr6, to 9.0.0), lodash-es prototype pollution (GHSA-f23m-r3pf-42rh to 4.18.0, GHSA-xxjr-mmjv-4gpg to 4.17.23), morgan log forging (GHSA-4vj7-5mj6-jm8m to 1.11.0, GHSA-9f6g-j8ch-79g4 to 1.12.1, GHSA-jxfw-x594-9x9m to 1.12.0), notevil sandbox escape (GHSA-8g4m-cjm2-96wq, no fix yet), and [email protected] with nine moderate XSS, input validation and information exposure advisories (GHSA-3j7m-hmh3-9jmp, GHSA-cv27-6wvh-8x7j, GHSA-jxwj-j7wr-gfrw, GHSA-mjxr-4v3x-q3m4, GHSA-qhxp-v273-g94h, GHSA-rjqq-98f6-6j3r, GHSA-rm97-x556-q36h, GHSA-x3q4-9hxx-gx8m, GHSA-xc6g-ggrc-qq4r; upgrade to 2.18.0 or later covers them). Low: [email protected] silently disables size limits with an invalid limit value (GHSA-v422-hmwv-36x6, upgrade to 1.20.6).
Developer-only tools with known flaws (lower priority housekeeping) — @cyclonedx/[email protected] has a high-severity shell injection on Windows via the --workspace argument (GHSA-q69g-4hcv-6jg4, upgrade to 6.0.0). [email protected] has a moderate path traversal / arbitrary file read (GHSA-82fw-gwwq-j7x9, upgrade to 4.1.11). Both are build or testing tools that never ship to real users.
💡 Key concept: Your app is only as safe as the oldest borrowed code inside it, so regular audits and updates matter as much as your own code.
Verified Findings
Show all 30 ↓
Private Key File — infrastructure/terraform/networking.tf line 171
A private key is stored in your code. This is the digital equivalent of committing your house keys.
Private Key File — lib/insecurity.ts line 21
A private key is stored in your code. This is the digital equivalent of committing your house keys.
Private Key File — terraform/networking.tf line 171
A private key is stored in your code. This is the digital equivalent of committing your house keys.
JSON Web Token — frontend/src/app/app.guard.spec.ts line 46
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — frontend/src/app/last-login-ip/last-login-ip.component.spec.ts line 72
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — frontend/src/app/last-login-ip/last-login-ip.component.spec.ts line 78
A login token is written into your code. These are meant to be short-lived, not stored.
Generic Hardcoded Secret — test/api/2fa.test.ts line 42
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 66
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 85
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 104
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 123
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 142
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 182
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 214
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 240
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 257
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 266
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 292
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 317
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 352
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 379
Something that looks like a password or key is typed directly into your code.
Generic Hardcoded Secret — test/api/2fa.test.ts line 412
Something that looks like a password or key is typed directly into your code.
JSON Web Token — test/api/user.test.ts line 280
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — test/cypress/e2e/forgedJwt.spec.ts line 22
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — test/cypress/e2e/forgedJwt.spec.ts line 38
A login token is written into your code. These are meant to be short-lived, not stored.
Generic Hardcoded Secret — test/cypress/e2e/totpSetup.spec.ts line 7
Something that looks like a password or key is typed directly into your code.
JSON Web Token — test/server/currentUser.unit.test.ts line 31
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — test/server/currentUser.unit.test.ts line 34
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — test/server/verify.unit.test.ts line 308
A login token is written into your code. These are meant to be short-lived, not stored.
JSON Web Token — test/server/verify.unit.test.ts line 316
A login token is written into your code. These are meant to be short-lived, not stored.
Check it yourself · about 5 minutes
Is a secret saved in your GitHub repository?
  1. Open your repository on GitHub.
  2. Open the file and line that Verilay named in this report.
  3. Decide whether the value is real, or an example or placeholder.
✓ It's an example, a placeholder, or a value that is meant to be public.
⚠ It's a real key. Make a new key in the service's dashboard and switch to it. Deleting the file isn't enough, because the old key stays in the repository's history.
Never paste a real key into a chat, an email or a website while checking.
Is that login token real?
  1. Open the file Verilay named, on GitHub.
  2. Decide whether the value is a test or example, or came from a real logged-in account.
✓ It's a test or example value.
⚠ It's from a real account. Remove it, sign that account out everywhere and change its password.
Never paste a real token into a chat, an email or a website while checking.
Only try this on your own app, with test accounts. Never on someone else's app.
Worth A Look
A pointer, not a confirmed problem. It does not change your grade.
Worth a look: An old hash (MD5 or SHA-1) is used, and this app also handles passwords
lib/insecurity.ts line 41 hashes with MD5 or SHA-1, which are broken for protecting passwords. The app also appears to hash or check passwords (models/user.ts line 73). Verilay can't tell whether the two are connected, so this is a pointer to check, not a confirmed problem.
What to do: Check what that hash is used for. If it protects passwords or login tokens, replace it with bcrypt, scrypt or argon2. If it is only a cache label, you can ignore this.
Check it yourself · about 5 minutes
What is that old hash used for?
  1. Open the file Verilay named, on GitHub, and read the few lines around that hash.
  2. Ask your AI builder: "Is this hash used for passwords or login tokens anywhere? Review and advise, don't change anything yet."
  3. Read its answer.
✓ It's only used for labels, such as cache keys or file names.
⚠ It protects passwords or login tokens. Move to bcrypt, scrypt or argon2.
If you use a login service such as Supabase Auth or Clerk, your app may not store passwords at all.
Only try this on your own app, with test accounts. Never on someone else's app.
🔧Not sure how to fix these? A Verilay Self Fix session walks you through each fix with your AI builder, checks its replies, and rescans to confirm. Coming soon. Get early access →
Scores may vary slightly between runs as findings are AI-generated. A meaningful improvement (e.g. C → B) after applying fixes indicates real progress.

AI disclaimer: Analysis is generated by Claude AI and may contain false positives or miss issues. Always verify findings with your AI builder before making changes. This is not a professional security audit. See full terms.