[Sharing] cc-playbook — A Full-Lifecycle Harness Engineering System for Claude Code

What is cc-playbook?

A production-grade harness engineering playbook for Claude Code that transforms Claude from a capable model into a reliable, full-lifecycle development partner.

Agent = Model + Harness. The model is powerful but directionless. The harness channels that power into consistent, high-quality results.

Repository

https://github.com/lexxx837/cc-playbook

What's inside?

| Component | Count | Description |
|-----------|-------|-------------|
| Workflows | 16 | Full lifecycle: PRD → Spec → Init → Develop → Test → Security Audit → Tech Debt → Retrospective |
| Guides | 13 | Architecture, testing, security, frontend motion, third-party integration, etc. |
| CLAUDE.md Templates | 9 | Project-type-specific configs with built-in quality gates |
| Project Types | 11 | Backend, Full-Stack (Python/Node), Frontend, Mobile, Desktop, Agent Teams, etc. |
| Design System | 1 | Production-ready glassmorphism component library |
| Project-Level Skill Skeletons | 3 | Extend generic harness into project-specific domain knowledge |

Key Design Decisions

1. Deterministic Routing over Prompt Guessing

Instead of relying on the model to "figure out" the right approach, cc-playbook uses a decision tree:

Project Type (A-K) × Scale (small/medium/large) × Mode (product/tech)
    → Specific Workflow → Specific Template → Phase-by-phase execution with gates

2. Two-Layer Root Cause Analysis

Every bug analysis must answer:

1. What went wrong in the code?
2. Why was the code written this way? (process/methodology layer)

3. Project-Level Skill Framework

Generic harness has a ceiling. The skeleton framework lets you create project-specific guardians:

• backend-guardian — Validates API contracts before changes
• ui-consistency — Enforces design token usage
• stability-keeper — Prevents "fix A, break B" regressions

4. Self-Maintaining System

MAINTENANCE.md defines 8 update triggers with checklists to prevent harness rot over time.

Origin

Born from real-world project experience: MVP takes 1 day, but polishing takes 10 days — because the harness wasn't designed for sustained iteration. This playbook fills that gap.

Key lessons encoded from multiple project failures (architecture pivots, feature silos, UX fragmentation) into anti-patterns and quality gates.

Quick Start

# Clone and copy to your Claude config
git clone https://github.com/lexxx837/cc-playbook.git
cp -r cc-playbook ~/.claude/cc-playbook

Then add the routing instruction to your ~/.claude/CLAUDE.md (see README for details).

---

Happy to answer any questions or discuss the design decisions. PRs and feedback welcome!