Operator · Agentic Pentest

From domain to root.

Point Operator at a domain. It attacks, exploits, and hands you a working proof — live.

139Proven-exploitable
1,730Findings
5Live targets
Operator — api.example.com +
operator.planckdefense.com/scans

api.example.com

EXPLOITING FULL_VALIDATE RUNNING

API agent — 32 operations

›_operator console · 0 events
Working
Ask the operatorlive · steering

01 The proof

Open one. There’s the shell.

Every critical ships a working PoC — CVSS, the exact request, and the uid=0(root) it returned.

operator.planckdefense.com / findings / rce-eval
CRITICALA03NEWUnauthenticated Remote Code Execution via Python eval() in /eval endpoint

Confirmed by the AI exploitation agent with a working proof-of-concept.

CVSS 3.1
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
Confidence
✓ AI-validated
Frameworks
A03CWE-78
Remediation

Never pass user input to Python eval(). Replace with a safe parser/whitelist of allowed operations, or remove the endpoint entirely. If arbitrary evaluation is required, sandbox it in an isolated process with no OS access.

RequestGETCOPY
GET /eval?s=1337*1007 HTTP/1.1 Host: example.com:9000 User-Agent: aipentest-scanner/0.1 (+authorized security testing) --- next request --- GET /eval?s=__import__('os').popen('id').read() HTTP/1.1 Host: example.com:9000
Response200COPY
{ "message": "Evaluation result: 1346359" } --- next response --- HTTP/1.1 200 · server: nginx/1.31.4 content-type: application/json { "message": "Evaluation result: uid=0(root) gid=0(root) groups=0(root)\n" }
References
cwe.mitre.org/data/definitions/95.htmlowasp.org/…/Code_InjectionCWE-78: OS Command InjectionPortSwigger: OS command injection
↻ Re-test→ ReplaySTATUS open ▾
Ask AI about this finding

Ask anything about this finding — impact, exploitation, remediation, how to verify a fix.

e.g. how do I fix this? is it exploitable?Ask

02 The impact

One chain. Login to root.

The exact path an attacker walks from login to full compromise — mapped to MITRE ATT&CK.

operator.planckdefense.com / attack-path
Proven chain → business impact
1
TA0006 · ACCESS
Unauthenticated SQLi in /tokens login
2
TA0006 · CREDENTIAL
Forge auth token via md5(ctime)
3
TA0004 · PRIV-ESC
Read any user’s data via X-Auth SQLi
4
TA0002 · EXECUTION
RCE via OS command injection
5
TA0040 · IMPACT
root · full database dump
Credential AccessPrivilege EscalationExecutionImpact

03 Command center

Your exposure, as it builds.

A live risk score, the severity split, OWASP/MITRE coverage and verification confidence — the whole picture assembling itself as you scroll.

Security Overview
Attack surface intelligence
live

04 Deliver

Proof, where you already work.

Straight to Jira, GitHub and Slack — deduped by fingerprint. Export CSV, JSON, SARIF, PDF.

operator.planckdefense.com / integrations
Push findings to your stack
JiraNOT SET

One issue per new finding, deduped.

Connect
GitHubNOT SET

Issue per finding, by severity.

Connect
SlackNOT SET

Scan summary when a scan finishes.

Connect
WebhookNOT SET

POST findings JSON to SIEM / SOAR.

Connect

Ready when you are

Know what’s exploitable
before they do.

Point Operator at a domain. Get proof — not a to-do list.