fyi, If you commented **ARCHITECT**, this is the resource I was talking about. Also found in our resources page here Most AI coding projects go sideways for the same reason: the agent starts writing code before the important decisions are settled. This skill separates the work into two roles. The **Architect** inspects the project, chooses the approach, defines the interfaces and guardrails, and creates a build packet. The **Builder** gets one bounded task at a time, implements it, verifies it, and sends the result back for review. You can use your strongest model for architecture and let faster or less expensive agents handle the implementation. The skill works with Claude Code, Codex, and other AI tools that accept project or custom instructions. Inside the resource you will get: - the full Architect skill - simple installation instructions - a reusable Builder handoff format - acceptance and review checks so "done" means verified Download it here: https://github.com/Rlegaspi562/agent-skills/tree/main/architect Try it with: **"Architect this feature before any code is written."**