Penetration Testing
AI-powered black-box pen testing
Try asking
About
Penetration Testing is a black-box web application security testing tool that crawls your app's attack surface, scans for misconfigurations, and tests for the most common and damaging vulnerability classes — all powered by AI-generated payloads. It produces a structured report with severity ratings, proof-of-concept evidence, and remediation guidance.
The tool follows the standard penetration testing workflow: reconnaissance first to map endpoints and input surfaces, then targeted testing for injection, XSS, authentication bypass, and SSRF. AI reasoning is used to generate context-aware payloads rather than generic pattern matching, which means it finds vulnerabilities that signature-based scanners miss.
Important: Only use this tool on systems you own or have explicit written permission to test. Unauthorised security testing is illegal.
What you can do
- Crawl any web app to map its full attack surface — endpoints, forms, auth flows, tech stack
- Scan for security headers, exposed paths, server disclosure, and TLS issues
- Test for SQL injection, NoSQL injection, and command injection with AI-generated payloads
- Test for reflected and stored Cross-Site Scripting in HTML, attribute, and JavaScript contexts
- Test for authentication bypass, IDOR, and privilege escalation
- Test URL-accepting parameters for Server-Side Request Forgery
- Generate a structured report with executive summary, severity breakdown, and remediation steps
Who it's for
Security engineers, developers, bug bounty hunters, and DevSecOps teams who need to validate their applications against real attack vectors before release or as part of ongoing security reviews.
How to use it
- Start with recon to map the attack surface — this returns a session object for all subsequent skills
- Run scan_vulnerabilities to check headers, exposed paths, and TLS configuration
- Use test_injection, test_xss, test_auth, and test_ssrf — each requires authorized: true
- Finish with generate_report to compile all findings into a readable report
Getting started
No setup required — the tool runs on the platform's AI models by default.
Crawl a target URL to map its attack surface
Run vulnerability scans against the target including security header…
AI-powered injection vulnerability testing
AI-powered Cross-Site Scripting testing
Crawl a target URL to map its attack surface. Discovers endpoints, forms, input parameters, authentication flows, and tech stack. Returns a session that subsequent test skills use.
Run vulnerability scans against the target including security header checks, exposed path detection, server version disclosure, and TLS analysis. Uses nuclei when available.
AI-powered injection vulnerability testing. Uses LLM reasoning to generate context-aware SQL, NoSQL, and command injection payloads, execute them, and analyze responses. Requires authorization.
AI-powered Cross-Site Scripting testing. Generates context-aware XSS payloads for HTML, attribute, and JavaScript contexts. Tests both reflected and stored XSS. Requires authorization.
AI-powered authentication and authorization testing. Tests for auth bypass, IDOR, privilege escalation, and session handling issues. Optionally accepts test credentials. Requires authorization.
AI-powered Server-Side Request Forgery testing. Tests URL-accepting parameters for internal network access, cloud metadata exposure, and redirect-based SSRF. Requires authorization.
Compile all findings from a scan session into a structured penetration test report. Includes executive summary, severity distribution, PoC evidence, and remediation recommendations.
- Added subtitle, expanded description, and agent instructions
- Initial release
Related
Questions
Do I need permission before running a test?
Yes. Only test systems you own or have explicit permission to assess.
What should I run first?
Start with `recon`. It maps the attack surface and creates the session used by the later tests.
What kinds of issues does it look for?
It covers injection, XSS, SSRF, authentication problems, misconfigurations, and related web app weaknesses.
How do I get a readable report?
Use `generate_report` at the end of the workflow. You can choose markdown for reading or JSON for structured output.