aleksacom/claude-code-orchestration-agents

Claude Code Orchestration Agents is a pre-built set of five sub-agents and a /change slash command designed to add structure and discipline to projects built with Claude Code, Anthropic's AI coding tool. Without guardrails, AI coding assistants can make sweeping changes, silently bake in architectural decisions, and leave no record of why things were done a certain way. This project addresses that by inserting checkpoints at every stage of a code change. The five agents each handle a specific role. The cm-planner acts as an architectural authority that plans the scope of every change before any code is touched and flags decisions that need your input. The cm-decider routes questions to either "decide now" or "auto-resolve" based on a set of trigger criteria. The reviewer examines every staged diff before it is committed and categorizes findings as blocking, medium, or minor. The conventions-audit agent enforces documentation conventions when decision files are touched or a phase of work closes. The test-author adds tests for any new behavior added to monitored source paths. The /change slash command orchestrates all of them in order: plan, verify, implement, review, audit, test, and then commit with a conventional commit message. To install, you copy the .claude/ and docs/ directories from this repository into your project root, update your project's CLAUDE.md with routing rules, and restart Claude Code. The README estimates about 15 minutes of setup time. Each agent is a plain markdown file you can edit to match your project's specific patterns. It requires Claude Code and an existing git-initialized project.