For builders without a security team

Let the machines try to break your app.

Name: Guardian Script
Price: 50

AI agents attack your app. From $100. Start with a free public review by email. No active testing.

AI agents pursue a goal, connect clues, and change tactics when blocked. We test what they could do to your app—and show you what to fix first.

Submitting a URL requests a review. Active testing starts only after we verify your authority and agree the scope.

Explicit authorizationHuman-controlled scopeReproducible evidence

Drill Me mascot proudly celebrates a server cluster surviving a strength test.

DRILL ME / DRILL ME / DRILL MEPERMISSION FIRSTREALISTIC, NOT RECKLESSPROOF OVER PANICFIXES YOU CAN SHIP

An example, not a customer claim

This is what useful proof looks like.

A teaching example of a paid drill report. All accounts, orders and actions below are synthetic. The free review does not include active testing.

Illustrative scenario — not a recorded runPermission boundary crossed

One test user could open another test user’s order.

The goal
Could test account A reach an order belonging to test account B?
Check the boundary
The agent starts with the order view available to account A.
A defense holds
That page blocks access to the other account’s order.
Change the approach
The agent checks another order-view route inside the same approved test scope.
The illustrated outcome
This second route returns account B’s synthetic order. The permission check is inconsistent.

See the synthetic example

Expected behavior: Account A can view only its own orders. Every route must enforce the same permission check.
Illustrated result: The second route exposes a synthetic order belonging to account B. This is made-up example data, not captured customer evidence.
What to fix first: Check on the server that every requested order belongs to the signed-in user. Add a test that rejects access from another account.
Assessment limit: This example covers one workflow. It is not a security verdict on the whole product.

Your paid report, in plain language

Outcome and impact: what the agent could do.
Evidence and first fixes: what your developer can verify and change.
Scope and limits: what we tested and which defenses held.

Start with the free review

A different threat model

Your scanner found nothing. What could an agent do?

A scan helps find known issues. A scoped drill tests whether an agent can turn a weakness into a concrete outcome, such as crossing a user-permission boundary.

They adapt

A scan checks for known patterns and configuration issues.

Agents observe responses, change tactics, and retry in real time.

They chain context

An isolated finding may not explain the risk to your product.

Agents combine small clues across pages, tools, and workflows.

They can act

A scan gives you a useful list of issues to investigate.

An agent can navigate, persuade, call tools, and pursue a goal.

A human-led penetration test can cover broader expert investigation and formal reporting. Drill Me is a focused agent-driven exercise; it does not replace a compliance audit.

Is this for you?

Built an app? Test what it lets an agent do.

Especially useful for small teams shipping custom products without a dedicated security team.

A good fit

Indie SaaS, custom apps and products built with AI coding tools
Client products, mobile backends and APIs—with permission

Start free. Go deeper when ready.

Choose the question. Let agents test it.

Start with public observations, or choose a paid drill to test an agreed goal. We confirm scope and timing before any testing.

FREE START

00 / PROTECTION SNAPSHOT

Free Gap Analysis

A review of publicly visible information, delivered by email. We separate what we can observe from questions that need testing.

^$0email delivery
no active testing

Up to three priority observations
Questions for an agent-driven drill
A practical next step

What a free review can look like

Illustrative example, not a review of your app.

Public fact: The site describes shared workspaces and different user roles.
Question to investigate: Could an agent move beyond the permissions of its test role?
Not visible publicly: Whether private data and server-side permission checks are protected.
Suggested next step: Agree a role-boundary scenario using synthetic test accounts.

01 / FAST ASSESSMENT

Simple Drill

Start with one specific question about what an agent could do to your app.

^$100one-time
~3 hours of testing

Goal: Test one focused agent-driven scenario.
Scope: One agreed target and one scenario.
Result: Findings, evidence, safe reproduction steps and prioritized fixes.

Start a Simple Drill

DEEPER PRESSURE TEST

02 / ADVANCED ASSESSMENT

Hard Drill

For a more complex attack chain across agreed product workflows.

^$500one-time
several days of testing

Goal: Test a more complex, multi-stage attack path.
Scope: Agreed workflows, methods and boundaries for several days of testing.
Result: Everything in Simple Drill, plus analysis of the attack chain.

Methods may include approved social engineering and frontier-model simulation.

Start a Hard Drill

Times describe approximate testing effort, not report delivery. We agree the schedule, total price and any follow-up before you commit.

Need a custom scope? Tell us what you need to protect.

The process

Your permission. A clear boundary. Then the drill.

A URL is the starting point, not permission to attack. We agree the rules with you before any active testing.

Verify authority

We confirm that you own the product or have permission to authorize testing, and identify a responsible contact.

Agree the scope

Together we define the target, environment, test accounts, allowed methods, exclusions, and stop conditions.

Run and record

Agents work inside the approved scope. We keep a timestamped record of actions and the evidence behind each finding.

Stop at the agreed limit

We stop at the agreed evidence threshold or when you ask. You receive the results, the limits of the test, and prioritized fixes.

Real pressure. Hard boundaries.

Written permission and scope come first. No data theft or destructive actions. We use only the access needed for evidence, and you can stop the drill at any time.

Optional after your drill

Keep watch after the drill.

Ask about Guardian Script for agentic traffic monitoring, alerts and common-pattern blocking. We’ll discuss whether it fits your product after the drill.

$50/ month

Discuss Guardian after your drill

Questions, answered

Know the boundaries before we begin.

What is included in the free analysis?

We review publicly accessible information and email up to three priority observations, open questions and a practical next step. We distinguish visible facts from hypotheses. Private workflows and vulnerabilities are not verified by this free review.

What happens after I send a request?

We reply by email to discuss your product and confirm the scope, timing and price where applicable. A URL and your main concern are enough to start; we can agree the scenario together. Sending a request does not start a test or a payment.

Is this a penetration test?

It is a focused adversarial assessment for agent-driven threats. It can complement a traditional penetration test, but it concentrates on adaptive behavior, workflow abuse, tool use, and realistic attack chains.

Will you access or take our data?

No. We do not steal data, perform destructive actions, or move outside the written scope. Evidence is captured with the minimum access needed to demonstrate a finding, and the drill stops at agreed boundaries.

What do you need to begin?

A target URL, proof that you are authorized to test it, a primary contact, and clear scope constraints. Hard Drills may also need approved test personas, communication channels, and staging accounts.

What does “social engineering” include?

Only scenarios you approve in advance, using named test personas and agreed channels. We do not contact customers, employees, or third parties without explicit written authorization.

What if you find nothing?

You still receive what we tested, the defenses that held, and the assessment limits. You pay for the agreed testing work, not a guaranteed vulnerability. No confirmed finding does not mean the app is risk-free.

Does Guardian Script replace a WAF or security team?

No. It adds an agentic-traffic signal and response layer. It is designed to strengthen existing controls by detecting probable automation, alerting your team, and blocking common attack patterns.

Your systems will be tested

Invite the attack. Control the outcome.

Send your URL and the outcome you want to prevent. We’ll agree the scenario, scope and next steps by email.

Book a $100 drill Review our boundaries