Simple Claude Orchestrator for Product Engineering
SCOPE turns Claude Code into a structured product engineering environment. It gives you slash commands that take a product from PRD to implemented, tested, and audited epics — with approval gates at every step.
PRD draft → /prd_refine → /prd_breakdown → /epic_refine → /implement → merge
Already have code but no docs? Use /re_documentation to reverse engineer the product and architecture documentation from your existing codebase.
No custom tooling. No MCP servers. Just Claude Code slash commands, agents, skills, and documentation templates.
/prd_refine— Interactively refine a product requirements document using a checklist-driven approach/prd_breakdown— Break the PRD into implementable epics with architecture and dependency analysis/epic_refine— Refine each epic through 4 approval gates: business validation, architecture design, spec generation, story breakdown with executable Python contracts/implement— Implement an epic story-by-story in a git worktree. Developer writes code and tests. Architect scaffolds shared modules first/implement_tdd— Same as above, but SDET writes tests first, then developer implements to make them pass/audit_epic— Audit the implementation against architecture, acceptance criteria, and specs. Auto-generates fix stories for all findings/sync_product— Update product documentation when implementation reveals scope changes
/re_documentation— Reverse engineer product and architecture documentation from an existing codebase. Two agents scan your code, interview you about decisions and rationale, then generate 24 documentation files (9 product + 15 architecture)
Every command has user approval gates. Nothing is merged without your sign-off.
SCOPE uses Claude Code's built-in features:
- Slash commands (
.claude/commands/) define multi-phase workflows with approval gates - Agent definitions (
.claude/agents/) give Claude specialized personas (architect, developer, SDET, product owner, reverse-engineer-po, reverse-engineer-architect) - Skills (
.claude/skills/) provide documentation templates (Arc42+C4 for technical, Atlassian Blueprint for product) - TaskCreate/TaskUpdate manage story dependencies and sequencing
- Git worktrees isolate implementation from the main branch
No external dependencies. No install scripts. No state files.
git clone https://github.com/PatD42/scope.git
cd scopeInstall to a project (commands available only in that project):
./install.sh /path/to/your-projectInstall to user directory (commands available in all projects):
./install.sh --userInstall to current directory (default):
./install.shThe script copies commands, agents, skills, and a config template into .claude/. For project installs, it also creates .scope/config.yaml — edit it to set your project name and tracking preferences.
Starting from a PRD:
1. Write a PRD draft (or bring an existing one)
2. /prd_refine → Refine it interactively
3. /prd_breakdown → Get epics with dependencies
4. /epic_refine EPIC-001 → Refine the first epic (4 approval gates)
5. /implement EPIC-001 → Implement it story-by-story in a worktree
6. Review and merge the worktree when satisfied
Starting from existing code:
1. /re_documentation → Reverse engineer product + architecture docs
2. /prd_refine → Refine/extend the PRD for new features
3. /prd_breakdown → Break into epics
4. Continue as above
After installation, your project will generate this structure as you work:
your-project/
├── .claude/
│ ├── commands/ # Slash commands
│ ├── agents/ # Agent definitions
│ └── skills/ # Documentation templates
├── .scope/
│ └── config.yaml # Project configuration
├── docs/
│ ├── product/ # Product docs (strategy, definition, decisions)
│ ├── architecture/ # Technical docs (Arc42 sections 01-13)
│ ├── epics/{epic-id}/ # Per-epic docs (acceptance criteria, file plans, ADRs)
│ └── releases/ # Release documentation
├── wip/
│ └── {epic-id}/ # Git worktree per epic (implementation happens here)
└── src/ # Your application code
Contract-first development — /epic_refine produces a contracts.py with Python Protocol classes. mypy --strict verifies every story implements the agreed interfaces. No more integration failures hidden by mocks.
File plan intent — Every file in a story has a 600-1200 char description covering WHAT, WHY, RESPONSIBILITIES, DEPENDENCIES, and RELATED MODULES. This is the source of truth for implementation.
Audit loop — After implementation, /audit_epic checks architecture compliance, acceptance criteria coverage, code quality, and stub detection. Fix stories are generated for all findings. Max 2 audit cycles, then escalate.
Git worktrees — Implementation happens in wip/{epic-id} on branch epic/{epic-id}. Main branch stays clean. You merge when satisfied.
- Claude Code CLI
- Git
MIT
