Experiment shows LLM session history can override skill-file rules even after correction
KhuyenTran16 · x · 2026-10-06
Khuyen Tran shares a hands-on experiment: skill files are supposed to give models stable rules, but after a model broke a rule once, it sometimes kept following the bad pattern later—even after correction. Suspecting session history outweighs the skill file, she tested it with a simple commit-message rule, changing only the earlier conversation history and checking whether the final answer still complied. The article covers the setup, results, and practical habits for making skill-file rules more reliable.
More from coding & agent
- Stack Overflow 2026 survey: 66% use AI coding tools, only 24% of firms have AI policies — pchandrasekar · 2026-10-06
- Shaders open-sources 200+ WebGPU effect components with free code export for React, Vue, Svelte, Solid — nicolascraske · 2026-10-06
- SOTA TV Podcast Launches With OpenHands Co-Founder and CMU Prof Graham Neubig — Madisonkanna · 2026-10-06
- DHH: To enjoy building with agents, you have to actually care about the software — csuwildcat · 2026-10-06
- Heavy Claude User on Gemini 4 Argon: Doesn't Beat Claude for Code, Antigravity Feels Alien — MicahBerkley · 2026-10-06
- ChatGPT Agent Answers Real Phone Calls With a $6 ESP32, Undercutting BPOs at $32/Month — jxnlco · 2026-10-06