Dev shares dual-model agent setup: Opus 5.5 for strategy, Sol 6.2 for implementation, plus free skills
PawelHuryn · x · 2026-11-19
Pawel Huryn details his current multi-agent coding workflow: Opus 5.5 (xhigh) handles judgment, common sense, strategy and orchestrating other agents, while Sol 6.2 (xhigh) runs implementation and code-review loops.
He open-sourced a free skills collection for Opus including /delegation, /codex-agents, /grok-agents, /code-review and /independent-review.
He also researched driving Google's Antigravity CLI (Neon) from other agents for a benchmark, but warns it violates Google's terms of use and could get accounts banned — the skill is ready if policy changes, but don't use it yet. A benchmark site compares score vs. API cost.
More from coding & agent
- ElevenLabs launches Scribe v3 Live Medical with 10,000-entry custom term lists — HeyAmit_ · 2026-11-19
- Developer on why personal AI agents still fail: 1Password auth, Mac control, mobile use — lucasmeijer · 2026-11-19
- Tencent's open-source Hunyuan Hy4 flash: 12B active params, 1M context, 50T tokens day one — HeyAmit_ · 2026-11-19
- Open-source /doomscroll skill lets your agent read 272 posts for $0.02 — msrivastav13 · 2026-11-19
- Testing tip: ban setTimeout in CI — use fake timers or wait on real events — DanielLockyer · 2026-11-19
- Vibe Design Tips launches free: 999 tips to make vibe-coded products feel polished — sujingshen · 2026-11-19