telus-labs/stagecraft

Stagecraft

Stagecraft is an orchestrator that runs your AI coding tool through a structured 18-stage pipeline. PM writes the brief. Principal designs. Specialists build their areas. Reviewers critique. QA tests. Each stage produces an artifact and a machine-readable gate. The next stage cannot start until the gate passes. The full run is on disk: auditable, resumable, not buried in a chat log.

Works across Claude Code, Codex CLI, Gemini CLI, and a generic no-host mode. One project, one config, one or more hosts. Different roles can run on different models — Claude for design, Codex for backend, Gemini for QA, Claude for review. The gate JSON is the seam.

devteam init --host claude-code        # one-time install in your project
devteam stage requirements --feature "Add SMS notification opt-in"
# (model writes brief + gate; hooks validate)
devteam next                           # → "▶️ run-stage design (stage-02)"
# … 16 more stages, then "🎉 pipeline-complete"

The CLI binary is devteam. Stagecraft is the project; devteam is what you type.

Table of contents

• First 30 minutes
• Documentation map
• Why "Stagecraft"?
• What this gives you
• Prerequisites
• Quick start
• What devteam init installs
• CLI reference
• Auditing an existing codebase
• Architecture in one diagram
• Repository layout
• Design decisions
• What's deferred
• Why this exists
• License

First 30 minutes

If you're evaluating Stagecraft, this is the fastest path to a working answer:

1. (5 min) Install the framework. git clone <this-repo> && cd stagecraft && npm install && npm link. Verify with devteam --help.
2. **(