AGT Suite · Security

AGT Shield

Most small businesses have never had their website or software looked at by anyone security-minded. Not because they don't care — because a security firm costs more than the website did. AGT Shield closes that gap: it inspects your live site, your source code, your database rules and every third-party library you depend on, then hands you one plain-English report saying what's wrong, how serious it is, and exactly how to fix it.

Scan Modules
4
Website, code, database rules, dependencies
Code Rules
30+
Python, JS/TS, PHP, Dart, C#, Java, Go
CVE Ecosystems
5
npm, PyPI, Dart, Composer, NuGet
Report Formats
4
HTML, PDF, Markdown, JSON
Automated Tests
64
Validated against ~28,000 files

Four modules, one report

Each module runs independently and in parallel. If one can't complete, the others still finish and the report says plainly which check didn't run — because a scanner that failed is not the same thing as a clean result.

Your live website

How your site looks to someone probing it from outside.

  • SSL certificate validity, expiry and strength
  • Outdated encryption still accepted (TLS 1.0/1.1)
  • Missing browser protections: HSTS, CSP, clickjacking defence
  • Login cookies that JavaScript can steal
  • Exposed files: .env, .git, backups, admin panels
  • Server software advertising its exact version

Your source code

Static analysis — the code is read, never executed.

  • Passwords and API keys hardcoded into files
  • SQL injection and command injection patterns
  • Cross-site scripting risks
  • Passwords stored with broken hashing
  • Security checks switched off (certificate validation, JWT signatures)
  • Debug mode left enabled in production

Your database rules

For Firebase and Supabase apps, these rules are the only thing between your data and the internet.

  • Firebase: collections readable or writable by anyone
  • Supabase: tables with Row Level Security off, or policies open to the public key
  • Temporary "test mode" rules never replaced
  • Data any signed-up account can reach — not just yours
  • The difference between a public key (safe by design) and an admin key (never expose)

Your dependencies

The third-party code your software is built on, checked against public vulnerability databases.

  • Known CVEs across five package ecosystems
  • Severity tuned to whether the flaw actually reaches production — not just its raw score
  • The exact version that fixes each issue
  • Cross-checked by two independent engines

From permission to report

STEP 01

Written authorisation

Nothing is scanned until you've named the assets and signed off. The permission is recorded in the report itself.

STEP 02

Four parallel scans

Website, code, database rules and dependencies are examined at the same time. Everything is passive and read-only.

STEP 03

Ranked and scored

Findings are de-duplicated, graded by severity, and scored out of 100 — so repeat instances of one mistake count as one mistake.

STEP 04

Explained in plain English

Each finding gets a business-level explanation of what an attacker could actually do, alongside the technical detail your developer needs.

STEP 05

Re-scan to verify

After fixes, run it again. The report tells you what's resolved and what's new since last time.

A report you can act on

Every finding names the exact file and line number, or the exact URL. Every finding carries a fix — not "improve your security posture", but the specific change to make.

SeverityWhat it means for you
CriticalTreat as an incident. Either exploitable right now, or credentials that should be assumed compromised.
HighA realistic path to compromise or data loss. Fix next.
MediumMeaningful hardening. Schedule into the next development cycle.
LowHygiene. Worth doing, not worth an emergency deploy.
InformationalObservations and confirmations. No action needed.

Findings are labelled by confidence. Anything detected by pattern matching rather than direct observation is marked tentative and counted at reduced weight, so you're never pushed into urgent work over something a developer hasn't confirmed.

Passive, read-only, and never without permission

Nothing is exploited

AGT Shield looks; it does not attack. It never submits credentials, never attempts to bypass a login, and never writes to your systems. Finding an open door is not the same as walking through it.

Consent is enforced in code

A host that isn't in the signed authorisation raises an error and the scan stops. It isn't a checkbox or a policy — the tool physically cannot scan an unauthorised target.

Secrets stay redacted

If a password or API key is found, the report shows just enough to locate it — never the full value. The report is safe to email.

Where a human is still required

Any security product that implies it makes you safe is selling you something. Here is what AGT Shield genuinely does not cover.

A clean report is not proof that a system is secure. It means the checks that ran came back clean. It cannot mean more than that, and nobody's tool can.

Business logic

Whether one customer can view another customer's order by changing a number in the URL. That takes a person who understands what your application is for.

Anything behind a login

Scanning stops at the front door. Authenticated areas need a different kind of engagement.

Penetration testing

No exploitation, no social engineering, no phishing simulation. If you need those, you need a specialist firm, and I'll tell you so.

The first thing it found was mine

The first production run of AGT Shield was against this very website. It found that the entire source repository — including every historical version of every file — was publicly downloadable, because of a single subtle pattern in a hosting config file. A pattern I had written, and had looked at many times without seeing.

It was verified by hand, fixed, and confirmed closed by re-scanning. Then it went out to client projects.

That's the honest case for this tool: config mistakes are invisible precisely because the person who made them believes they're correct. A second set of eyes that never gets tired and never assumes is worth having.

Now taking assessments

Find out what's actually exposed

If I built your website or system, I can run an assessment on it. If someone else built it, I can still scan it — I'll just need your written go-ahead first.

Request an assessment See the AGT Suite