War diary from LLM-assisted pentesting. Needed a vulnerable Eclipse Che instance for my Metasploit PR (CVE-2025-12548). Asked Grok 4 and Claude Sonnet 4.6 — both confidently wrong in different ways. Phantom CLI flags, CrashLoopBackOffs, invisible trailing newlines.
The fix came from plain old human debugging instinct, not a clever prompt.
https://payloadforge.io/i-asked-two-ais-to-help-me-set-up-a-test-environment-heres-what-actually-happened/
#Metasploit #OffSec #ExploitDev #LLM
Payload Forge
I Asked Two AIs to Help Me Set Up a Test Environment. Here's What Actually Happened.