ccg-workflow

CCG Workflow is a multi-model orchestration engine for Claude Code that coordinates Claude, OpenAI Codex, and Google Gemini across software development tasks from a single entry point: the `/ccg:go` command. When invoked, the engine reads the project's git status, tech stack, and file structure, classifies the task by type and complexity, then selects one of ten strategies ranging from `direct-fix` for single-file bugs to `full-collaborate` for complex multi-module features. That last strategy spawns parallel Agent Teams Builders with isolated file ownership, runs dual-model analysis with Codex and Gemini in parallel, and enforces HARD STOP checkpoints before implementation proceeds. A hook engine injects per-turn state into Claude's context via structured `<ccg-state>` blocks, which keeps the session coherent even after context compaction. Domain knowledge hooks auto-inject relevant reference files when messages mention topics like security or caching. The v3.0 rewrite consolidates what previously required 29 commands into 13. Developers working on medium-to-large codebases with Claude Code installed will find the most utility here.

What people ask about ccg-workflow