New AI Framework Turns Coding Agents Into Autonomous Hacking Tools, Outperforms Industry Benchmarks

A new multi-agent framework called T3MP3ST is transforming AI coding tools like Claude Code and Codex into autonomous hacking systems capable of running full cyberattack chains, achieving a record-breaking 90.1% score on a leading industry benchmark with fully reproducible results.

• T3MP3ST is a multi-agent offensive-security framework that transforms existing AI coding agents like Claude Code, Codex, and Hermes into autonomous red-teaming tools, running a full kill chain from recon to exploit to report with no new API keys required.
• The platform achieves a 90.1% pass@1 score on XBOW's 104-challenge benchmark suite, outperforming XBOW's self-reported 85%, and all benchmark numbers are fully reproducible from committed data via a single command: npm run verify-claims.
• T3MP3ST currently supports stable operations across web apps, CTFs, and embedded/IoT vulnerability pipelines, while cloud, mobile, binary reverse engineering, and full 8-operator swarm exploitation remain in active development or scaffolding stages.