Subagent Driven Development

This is a structured workflow for executing implementation plans by spinning up fresh Claude instances for each task, with a two-stage review process checking spec compliance first, then code quality. You stay in the same session but get clean context per task, which prevents the usual "wait, what was I doing?" pollution. The overhead is real (three subagents per task: implementer, spec reviewer, quality reviewer), but the theory is sound: catching spec drift early means fewer wasted cycles on beautiful code that solves the wrong problem. Best for plans with 3-8 independent tasks where you want systematic execution without switching to parallel sessions. The verbose process diagrams and prompt templates suggest this needs tooling support to not feel bureaucratic.

Install to Claude Code

npx -y skills add sickn33/antigravity-awesome-skills --skill subagent-driven-development --agent claude-code