Project Development

Before you spin up another LLM project, this skill makes you stop and evaluate whether your task actually fits the model's strengths. It walks you through task-model fit tables to avoid wasting days building on a bad foundation, then guides architecture decisions for batch pipelines, multi-agent systems, or interactive apps. Originally from muratcankoylan's context engineering collection and now maintained by flora131. The approach is practical: run your idea through the framework first, then build. It's passed most security audits except a Snyk warning. Best for the planning phase when you're still figuring out if agents are the right tool or if you should just write normal code.

Install to Claude Code

npx -y skills add flora131/atomic --skill project-development --agent claude-code