GPN CTF 2026 — LLM Harness Post-Mortem: Where Claude Code Got It Right (and Wrong)
Platform GPN CTF 2026 (kitctf) Difficulty Meta — harness post-mortem OS Tooling — Claude Code (Opus 4.x, 1M context), Bash sandbox, sub-agents Tags orchestrating Claude Code with parallel sub-agents, using scratch directories as harness cache, keeping main-thread context lean by routing bulk output through sub-agents, building 'kill the wrong plan' as a forcing function, recognising harness hallucinations in less-common ecosystems by grep-confirmation The LLM harness post-mortem isn’t about one challenge — it’s about the workflow that produced the other 18 writeups in the GPN CTF 2026 repository. What the harness around Claude Code actually looked like during a 24-hour CTF, what it was good at, where it embarrassed me, and which design choices I’d keep. This writeup is the standalone version of the meta-writeup submitted for the Best LLM Harness Writeup prize at GPN CTF 2026. ...