Three Man Team
A structured three-agent development process for Claude Code and other context-file-based tools, with an Architect that plans and deploys, a Builder that implements exactly the brief, and a Reviewer that catches what the Builder missed.
Three Man Team gives an AI coding assistant three agents with distinct jobs and clear handoffs, aimed at the failure modes of undisciplined AI coding: reading entire codebases for one function, adding unrequested features, drifting mid-task, and burning tokens on unnecessary work. The Architect plans and deploys, the Builder builds exactly what the brief specifies, and the Reviewer does not pass work that is not right.
It runs inside a single Claude Code session: the Architect (Arch) is the main agent, and when work is ready to build it spins up the Builder (Bob) as a subagent through Claude Code's Agent tool, then the Reviewer (Richard) the same way, so a user never has to open three separate windows. It works with Claude Code, VS Code, Cursor, and any AI tool that supports context files.
Setup generates a manifest.md at first-time setup as the single source of truth for an install's team names, role filenames, handoff directory, repository and branch, and a version registry drives update checks.