@busf4ctor@rez0__@Rhynorater While GPT 5.5 medium, made less mistake and just created the app with same vuln only low impact security issues like CSP header misconfig. So if u want to use Claude for making CTF, double check it might create more vulnerabilities than usual :)
I'm honored to share VulnRepro
Can new AI models spot vulnerabilities in heavily tested bug bounty targets?
I built a benchmark using recent writeups (past 2–3 years) to replicate real-world targets, URLs, and headers as closely as possible.
👉 https://t.co/VLyz9rEhSf (1/2)
@busf4ctor@rez0__@Rhynorater While i was making apps with Ai, i gave the writeup paths, params, url etc and when AI made them, i found that apps made with Claude (4.7,sonnet 4,6) have more vulnerabilities than GPT 5.5. I asked for xss and app made with 4 more vulns like IDOR , Broken Access Control