Search Psychevus

12 results available.
↑ ↓ to navigate ↵ to open · esc to close

Psychevus Verify

Built with AI. Checked by a human.

Independent security reviews for apps made with Lovable, Cursor, Claude Code, Bolt and v0. I sign in as two of your customers and try to cross the line between them, the way an attacker would. Then you get the evidence and a fix you can paste straight into your AI builder.

  • From $490, fixed
  • Report in 2 business days
  • No Medium finding, no fee

What I check

AI tools write the happy path. Attackers take the other one.

Apps built fast tend to fail in the same ten places. None of them shows up in a demo, and most can’t be seen by an automated scanner, because seeing them takes two accounts and a reason to cross between them.

  1. 01

    Database access rules

    Row Level Security or Firebase rules: can a signed-in user read or change another user’s rows by editing one request?

  2. 02

    Secrets in the browser

    Service keys, Stripe secrets and AI provider keys shipped in your JavaScript, your repository or your logs.

  3. 03

    Authorization on every endpoint

    API routes, server actions and edge functions that trust an ID, role or price sent by the client.

  4. 04

    Sign-in and sessions

    Password reset, magic links, OAuth redirects, account enumeration and sessions that never end.

  5. 05

    Payments

    Webhooks without signature checks, plans unlocked in the browser, and refunds or credits anyone can trigger.

  6. 06

    File storage

    Public buckets, guessable file paths, and uploads that can carry a script into your domain.

  7. 07

    AI features and agents

    Prompt injection through user content, tools that act without confirmation, and one user’s data in another’s context.

  8. 08

    Abuse and runaway cost

    Unlimited sign-ups, one-time codes and AI calls: the difference between a bot and a five-figure model bill.

  9. 09

    Configuration

    Security headers, CORS, public source maps, debug routes and admin pages reachable from the internet.

  10. 10

    Your email domain

    Whether anyone can send email as you (SPF, DKIM, DMARC), checked with the same tools this site publishes.

Check it yourself first

Free tools, no account. Then decide if you need me.

Fixed prices

Three ways to work together. No hourly meter.

Launch Check

$490Report in 2 business days

An app about to launch, or live with its first users.

  • All ten checks below, against your live or staging app
  • Two test accounts tried against each other: can one see or change the other’s data?
  • Your database rules read from a schema-only export (never your data)
  • Every finding reproduced, with the evidence and a severity you can act on
  • A paste-ready fix prompt for Cursor, Claude Code, Lovable or Bolt, plus a one-minute check that it worked

Paid on delivery of the report. If it finds nothing rated Medium or higher, there is no fee.

Book a Launch Check

Most chosen for apps with paying users

Deep Review

$1,900Report in 5 business days

Paying customers, teams and roles, sensitive data, AI agents with tools, or a customer asking for a security review.

  • Everything in the Launch Check
  • Code review of authentication, authorization and tenant isolation (read-only repository access)
  • AI features: prompt injection, tool permissions, and data crossing between users in model context
  • Payments end to end: webhooks, entitlements, plan and price tampering
  • Dependencies, CI and deploy settings, logging and incident readiness
  • A 45-minute walkthrough call and one retest within 30 days
  • A sealed attestation letter you can share with customers and investors

Half on booking, half on delivery. Your Launch Check fee is credited in full if you upgrade within 30 days.

Book a Deep Review

Fix & Verify

From $1,200Fixed quote after a review

You’d rather have the findings fixed than fix them yourself.

  • Fixes as small pull requests you can read and merge, each with a test that fails before and passes after
  • Database policies, server-side checks, webhook verification and key rotation done properly
  • A full retest and an updated, sealed attestation

Fixed price per milestone, quoted from the review’s findings. No hourly surprises.

Ask about Fix & Verify

What you receive

A short list of real problems. Each one proven, each one fixable.

Not a scanner export with 300 maybes. Every finding is reproduced by hand, ranked by what it would cost you, and comes with a fix prompt and a one-minute check you can run yourself. The sample report shows ten of them for a fictional invoicing app.

  • 2 critical
  • 3 high
  • 3 medium
  • 2 low
Read the full sample report
CriticalQB-02 · Payments

The Stripe webhook accepts unsigned events: anyone can upgrade to Pro for free

The function trusts whatever JSON it receives. It has to be deployed without Supabase’s JWT check so Stripe can call it, which leaves Stripe’s signature as the only proof an event is real, and the signature is never checked. One request upgrades any workspace to the paid plan.

Evidence · Reproduced
curl -X POST https://<ref>.supabase.co/functions/v1/stripe-webhook \
  -d '{"type":"checkout.session.completed",
       "data":{"object":{"metadata":{"workspace_id":"<B>"}}}}'

200 OK   → workspace B shows “Pro” without a payment
Paste into your AI builder

The stripe-webhook Edge Function parses the request body with req.json() and never verifies the Stripe signature. Change it to read the raw body with req.text(), verify it with stripe.webhooks.constructEventAsync using the stripe-signature header and STRIPE_WEBHOOK_SECRET, and return 400 if verification fails. Before changing a plan, retrieve the checkout session from Stripe and only upgrade when payment_status is 'paid', using the workspace stored on our side for that Stripe customer, not metadata from the event.

Check it yourself: Send the curl request above again. It must return 400 and the workspace must stay on its current plan.

How it works

From first message to sealed report in about a week.

  1. 01

    Tell me about the app

    Two minutes on the form below. I reply within one business day with a fixed price, a start date and any questions.

  2. 02

    Authorize and share access

    A one-page written authorization, a staging or live URL, two test accounts, and a schema-only export. Never your production data.

  3. 03

    Review

    I test the two accounts against each other and read the paths that decide who can do what, money and AI tools first.

  4. 04

    Report, fix, retest

    Findings with evidence and fix prompts, a walkthrough if you want one, then a retest and a sealed report you can share.

The fine print, up front

What a review is, and what it isn’t.

  • A review is evidence about one version of your app at one point in time. It is never a guarantee that no vulnerability exists, and no honest reviewer will say otherwise.
  • It is not a compliance certificate (SOC 2, ISO 27001, PCI DSS). Findings are referenced to the OWASP Top 10 (2021) and the OWASP Top 10 for LLM Applications, so they slot into one later.
  • I only test systems you own and authorize in writing. No denial-of-service, no social engineering, no destructive tests outside staging.
  • Your code and report stay confidential. NDA on request; access is removed and working notes deleted 30 days after delivery.

Who you’re hiring

Psychevus is me: an independent engineer who builds software and security tools, and publishes them so you can check the work before you pay for it. Try the 19 free security tools, the action gate and agent authority labs on AI tool permissions, or the signed proofs this site issues. The same standard applies to your report: you can verify it.

Questions

Before you ask.

How is this different from an automated security scanner?

Scanners can’t sign in as two different customers and try to read each other’s data, which is the most common serious flaw in AI-built apps. Every finding here is reproduced by hand and comes with evidence, so you get a short list you can act on, not hundreds of maybes.

I built it with Lovable or Bolt and don’t really read code. Is this for me?

Yes. Every finding comes with a fix prompt written for your AI builder, plus a one-minute check you can run yourself to confirm the fix worked. If you’d rather not touch it, Fix & Verify does it for you.

Do you need my source code?

Not for the Launch Check: two test accounts and a schema-only export of your database are enough. The Deep Review reads the code, so it needs read-only access to the repository, which you can revoke the day the report arrives.

Will you touch my production data?

No. Staging is preferred. If production is all there is, testing uses only accounts you create for it, and nothing destructive runs there.

What if you don’t find anything?

If a Launch Check finds nothing rated Medium or higher, you don’t pay. You still get the report.

Can I show the report to my customers or investors?

Yes. The Deep Review ends with an attestation letter: what was reviewed, when, and which findings were fixed and retested. Its fingerprint is timestamped and signed by psychevus.com, so anyone can check it hasn’t been edited since. It states what was reviewed; it never claims the app is “secure”.

How do I pay?

By invoice, in US dollars or euros. The Launch Check is paid on delivery; the Deep Review half on booking and half on delivery; Fix & Verify per milestone.

Who does the work?

I do, personally: an independent engineer who builds and publishes security tools, open APIs and an MCP server that you can inspect on this site. No subcontractors, and nothing is pasted into a third-party AI tool without your written agreement.

Your app next

Tell me about it. Get a fixed quote within a business day.

Two minutes. No call needed, no payment now. If a review isn’t the right thing for your app, I’ll say so.

Enable JavaScript to send this form, or write to hello@psychevus.com.

Which review?
Built with (pick any)

Fix & Verify: From $1,200. No payment now. You’ll get a fixed quote within one business day.

YOUR VISIT, YOUR CHOICE

Privacy, by intention.

You decide what stays on your device. You can change your choice at any time.

Essential

Saves your privacy choice. Administrator sign-in uses a secure session cookie.

Always on
Full privacy notice ↗