three-man-team

Why we included this project

If your usual workflow with a single AI coding session consists of constant nudging to stay on task, this project fixes the process rather than the prompt. It sets up three agents with separate jobs: the Architect plans work and owns the deploy, the Builder implements exactly what the brief asks for, and the Reviewer sends work back when it is not up to standard. All three run inside one Claude Code session, with clear handoffs, a per-project manifest as the single source of truth, and rules that stop agents from over-reading the codebase or spending tokens on calls nobody asked for. It ships as skill and context files you copy into a project, plus a setup script that handles naming and structure. Solo developers and small teams already committed to Claude Code who want guardrails against agent drift and runaway token spend get a pragmatic, configurable starting point.