CCFlow - Claude Code Workflow System

Version: 1.0 (Release) Status: ✅ Production Ready Last Updated: 2025-10-05

Overview

CCFlow is a comprehensive workflow system for Claude Code that provides structured development with TDD enforcement, memory bank context management, and intelligent agent-based task coordination.

Key Features:

🎯 Two-layer agent architecture (workflow coordination + implementation)
✅ TDD workflow with 100% GREEN gate (tests must pass)
📚 Memory bank system for persistent project context
🤖 12 commands under /cf: namespace
👥 9 agents (6 workflow + 3 hub) + specialist extensibility
📊 4-level complexity assessment for intelligent routing
🔄 Interactive modes for ambiguous requirements

Quick Start

1. Initialize Your Project

New Project

/cf:init MyProject

Existing Project (Auto-Import)

/cf:init MyProject
# Detects README.md, CLAUDE.md, package.json
# Offers to import existing documentation
# Pre-populates memory bank → validate → fill gaps

Force Fresh (Ignore Existing Docs)

/cf:init MyProject --force-fresh

What gets created:

memory-bank/ directory with 6 core files
Project context (imported from docs OR guided creation)
Foundation for structured development

Import Sources (if detected):

README.md → Executive Summary, Features, Problem Statement
CLAUDE.md → Tech Stack, Constraints, Patterns
package.json → Dependencies, Framework detection
Code structure → Architecture style, Component patterns

2. Customize Hub Agents (Important!)

Before first use, edit these 3 files to match your tech stack:

Testing Configuration (.claude/agents/testing/testEngineer.md):

Testing Framework: [Jest, Pytest, etc.]
Test Commands: [npm test, pytest, etc.]
Coverage Thresholds: [80%, 90%, etc.]

Development Configuration (.claude/agents/development/codeImplementer.md):

Tech Stack: [Node.js, Python, Go, etc.]
Coding Conventions: [ESLint, Black, etc.]
Project Patterns: [From systemPatterns.md]

UI Configuration (.claude/agents/ui/uiDeveloper.md):

UI Framework: [React, Vue, Angular, etc.]
Styling: [CSS-in-JS, TailwindCSS, etc.]
Component Conventions: [Naming, structure]

3. Start Building

# Create your first task
/cf:feature "implement user authentication"

# Plan it (if Level 2-4 complexity)
/cf:plan TASK-001 --interactive

# Implement with TDD
/cf:code TASK-001

# Save progress
/cf:checkpoint

Command Reference

Initialization (2)

Command	Purpose
`/cf:init [project-name]`	Bootstrap project with memory bank (auto-imports existing docs)
`/cf:init [project-name] --force-fresh`	Bootstrap with empty templates (ignore existing docs)
`/cf:sync`	Read-only memory bank status summary

Core Workflow (3)

Command	Purpose	Agent(s)
`/cf:feature [description]`	Entry point for new work	Assessor
`/cf:plan TASK-[ID] [--interactive]`	Plan Level 2-4 tasks	Architect, Product, Facilitator
`/cf:code TASK-[ID]`	TDD implementation with GREEN gate	testEngineer, Hub agents

Support (6)

Command	Purpose	Agent(s)
`/cf:review [task-id\|all]`	Quality assessment	Reviewer
`/cf:checkpoint [--message "..."]`	Save progress	Documentarian
`/cf:ask [question]`	Query memory bank	None (semantic search)
`/cf:context [--full]`	Load work context	None (file reads)
`/cf:status [--filter ...]`	Quick task overview	None (file reads)
`/cf:facilitate [topic] [--mode ...]`	Interactive refinement	Facilitator

Agent Management (1)

Command	Purpose
`/cf:create-specialist [domain] --type [type] --name [name]`	Create domain specialist

Workflow Patterns

Simple Task (Level 1)

/cf:feature "fix login button styling"
# → Assessor rates as Level 1
# → Direct to implementation

/cf:code TASK-001
# → testEngineer writes tests (RED)
# → uiDeveloper implements (GREEN)
# → Done in < 2 hours

Complex Feature (Level 3)

/cf:feature "implement order management system"
# → Assessor rates as Level 3
# → Requires planning

/cf:plan TASK-002 --interactive
# → Architect designs system
# → Product defines requirements
# → Facilitator refines through dialogue
# → Creates sub-tasks

/cf:code TASK-002-1  # First sub-task
# → TDD workflow
# → Specialists may be delegated to

/cf:checkpoint  # Save progress
/cf:review TASK-002-1  # Quality check

Daily Workflow

# Morning: Load context
/cf:context
/cf:status

# Work: Implement tasks
/cf:code TASK-[ID]

# Regular: Save progress
/cf:checkpoint  # Every 30-60 minutes

# End of day: Review and save
/cf:review all
/cf:checkpoint --message "End of day"

Agent Architecture

Workflow Agents (6)

Located in .claude/agents/workflow/:

Agent	Responsibilities
Assessor	Complexity evaluation, routing decisions, codebase scanning
Architect	System design, technical planning, architecture decisions
Product	Requirements, user experience, feature prioritization
Documentarian	Memory bank maintenance, checkpoint creation, knowledge preservation
Reviewer	Code quality, technical debt, progress evaluation
Facilitator	Interactive refinement, gap identification, collaborative exploration

Implementation Hub Agents (3)

Stack-agnostic coordinators that delegate to specialists:

Hub Agent	Domain	Delegates To
testEngineer	TDD coordination	Testing specialists
codeImplementer	Backend/business logic	Development specialists
uiDeveloper	Frontend/UI	UI specialists

Creating Specialists

# Create a database specialist
/cf:create-specialist database --type development --name databaseSpecialist

# Create a chart specialist for UI
/cf:create-specialist data-visualization --type ui --name chartSpecialist

# Create an API integration test specialist
/cf:create-specialist api-integration --type testing --name apiIntegrationSpecialist

Memory Bank System

6 Core Files

Located in memory-bank/:

File	Purpose	Updated By
projectbrief.md	Scope, objectives, constraints, decisions	Manual + Documentarian
productContext.md	Features, requirements, user flows	Product agent + Documentarian
systemPatterns.md	Architecture, patterns, conventions	Architect agent + Documentarian
activeContext.md	Current focus, recent changes, next steps	All agents + Documentarian
progress.md	Completed work, milestones, checkpoints, tech debt	Reviewer + Documentarian
tasks.md	Task tracking, status, complexity, sub-tasks	Assessor + various agents

Update Strategy

Hybrid Approach:

Auto-tracked changes during operations
Explicit commit via /cf:checkpoint
Agents update relevant files during workflow
Documentarian ensures cross-file consistency

TDD Workflow

Absolute Requirement: Tests must pass (100% GREEN gate)

RED → GREEN → REFACTOR

/cf:code TASK-001

# Step 1: RED Phase (testEngineer)
# - Write failing tests FIRST
# - Tests define expected behavior
# - No implementation yet

# Step 2: GREEN Phase (Hub Agent)
# - Implement to make tests pass
# - May delegate to specialists
# - Verify ALL tests pass

# Step 3: GREEN Gate Enforcement
# ✅ Tests pass → Update memory bank, mark complete
# ❌ Tests fail → Iterate (max 3 attempts)
# ❌ Still failing → STOP, report blocker

# Step 4: REFACTOR (Optional)
# - Improve code while maintaining GREEN
# - Tests must stay passing

GREEN Gate is Non-Negotiable: Implementation cannot be marked complete until all tests pass.

Complexity Levels

Level	Name	Criteria	Time	Routing
1	Quick Task	Simple, 1-2 files	< 2 hours	`/cf:code` (direct)
2	Simple Enhancement	2-5 files, clear scope	2-6 hours	`/cf:plan` required
3	Intermediate Feature	5-15 files, moderate complexity	1-3 days	`/cf:plan` required
4	Complex Project	15+ files, high complexity	3+ days	`/cf:plan` required

Assessor agent analyzes:

Keywords (implementation indicators)
Scope (files/components affected)
Risk (dependencies, unknowns)
Effort (estimated time)

Interactive Modes

Facilitator Modes

Exploration (--mode explore):

Open-ended discovery
Ambiguous requirements
Early-stage ideation

Refinement (--mode refine, default):

Improve existing plans
Fill gaps in specifications
Resolve ambiguities

Validation (--mode validate):

Pre-implementation check
Catch assumptions
Edge case discovery

Example Usage

# Explore a new feature idea
/cf:facilitate "notification preferences" --mode explore

# Refine an existing plan
/cf:facilitate TASK-042 --mode refine

# Validate before implementation
/cf:facilitate TASK-035 --mode validate

File Structure

your-project/
├── memory-bank/
│   ├── projectbrief.md
│   ├── productContext.md
│   ├── systemPatterns.md
│   ├── activeContext.md
│   ├── progress.md
│   └── tasks.md
├── .claude/
│   ├── agents/
│   │   ├── workflow/          # 6 workflow agents
│   │   ├── testing/           # testEngineer + specialists/
│   │   ├── development/       # codeImplementer + specialists/
│   │   └── ui/                # uiDeveloper + specialists/
│   └── commands/              # 12 command files
└── [your source code...]

Best Practices

Regular Workflow

✅ DO:

Run /cf:checkpoint every 30-60 minutes
Use /cf:context at session start
Run /cf:review before marking tasks complete
Create specialists when patterns emerge (3+ delegations)
Follow TDD workflow (tests first, always)
Use /cf:status for quick orientation

❌ DON'T:

Skip writing tests (TDD is enforced)
Mark tasks complete with failing tests
Let more than 1 hour pass without checkpoint
Create specialists prematurely (wait for patterns)
Ignore technical debt from /cf:review

Planning Strategy

Level 1 (Quick tasks):

Go straight to /cf:code
Keep it simple and fast

Level 2-4 (Complex tasks):

Always use /cf:plan first
Use --interactive for ambiguous requirements
Break into sub-tasks
Validate plan with /cf:facilitate --mode validate

Quality Management

Run /cf:review all weekly
Address high-priority tech debt promptly
Track quality trends in progress.md
Update systemPatterns.md as patterns evolve

Documentation

User Guides

Getting Started - 5-minute quick start
Command Reference - All 12 commands with examples
Agent Reference - Understanding the 9 agents
Workflow Patterns - Common development patterns

System Documentation

Architecture - Core architectural concepts
Validation - Quality standards and methodology
Extending CCFlow - Customization and extension guide

Technical Specifications

Workflow System - Complete workflow specification
TDD System - TDD workflow and agent system
Initialization - Project init process
Facilitator Patterns - Question templates
Creative Sessions - Deep exploration process

Implementation Files

All commands and agents fully documented in .claude/:

Commands: .claude/commands/cf/ (12 commands)
Agents: .claude/agents/ (9 agents)
Templates: .claude/templates/ (6 memory bank templates)

Validation & Quality

Status: ✅ Production Ready

✅ Command consistency: 100%
✅ Agent frontmatter: 9/9 valid
✅ Template consistency: 6/6 complete
✅ Integration points: 100% valid
✅ Cross-references: 100% correct
✅ Documentation: 100% complete
✅ Error handling: 30+ cases documented

Quality Metrics:

Total lines: ~20,000 across implementation files
Documentation coverage: 100%
Consistency score: 100%

See Validation Guide for methodology and standards.

Version History

v1.0 (2025-10-05) - Initial Release ✅

All 8 phases complete
12 commands implemented
9 agents created
6 memory bank templates
Complete validation passed
Production ready

License

This system is designed for use with Claude Code by Anthropic.

Support & Feedback

For issues, questions, or feedback:

Review documentation in .claude/commands/cf/ for command usage
Check memory-bank/ files for project context
Refer to agent files in .claude/agents/ for agent behavior

Built with Claude Code 🤖

Version: 1.0 (Release) Status: ✅ Production Ready Validation: ✅ All Checks Passed (12/12)

Name		Name	Last commit message	Last commit date
Latest commit History 7 Commits
.claude		.claude
docs		docs
CLAUDE.md		CLAUDE.md
README.md		README.md

ScaleLabs-Dev/CCFlow

Folders and files

Latest commit

History

Repository files navigation