🚀 Revolutionary AI-powered development workflow orchestration system ## 🔥 Core Innovations - **Document-State Separation**: Markdown for planning, JSON for execution state - **Progressive Complexity Management**: Level 0-2 adaptive workflow depth - **5-Agent Orchestration**: Specialized AI agents with context preservation - **Session-First Architecture**: Auto-discovery and state inheritance ## 🏗️ Key Features - Intelligent workflow orchestration (Simple/Medium/Complex patterns) - Real-time document-state synchronization with conflict resolution - Hierarchical task management with 3-level JSON structure - Gemini CLI integration with 12+ specialized templates - Comprehensive file output generation for all workflow commands ## 📦 Installation Remote one-liner installation: ``` iex (iwr -useb https://raw.githubusercontent.com/catlog22/Claude-CCW/main/install-remote.ps1) ``` ## 🎯 System Architecture 4-layer intelligent development architecture: 1. Command Layer - Smart routing and version management 2. Agent Layer - 5 specialized development agents 3. Workflow Layer - Gemini templates and task orchestration 4. Memory Layer - Distributed documentation and auto-sync 🤖 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>
9.1 KiB
Workflow Session Management Principles
Overview
This document provides complete technical implementation details for session state management, multi-session registry, command pre-execution protocol, and recovery mechanisms.
Multi-Session Architecture
Session Registry System
Lightweight Global Registry: .workflow/session_status.jsonl
The system supports multiple concurrent sessions with a single active session:
{"id":"WFS-oauth-integration","status":"paused","description":"OAuth2 authentication implementation","created":"2025-09-07T10:00:00Z","directory":".workflow/WFS-oauth-integration"}
{"id":"WFS-user-profile","status":"active","description":"User profile feature","created":"2025-09-07T11:00:00Z","directory":".workflow/WFS-user-profile"}
{"id":"WFS-bug-fix-123","status":"completed","description":"Fix login timeout issue","created":"2025-09-06T14:00:00Z","directory":".workflow/WFS-bug-fix-123"}
Registry Management:
- Single Active Rule: Only one session can have
status="active" - Automatic Registration: Sessions auto-register on creation
- Session Discovery: Commands query registry for active session context
- Context Inheritance: Active session provides default workspace and documents
Command Pre-execution Protocol
Universal Session Awareness: All commands automatically check for active session context before execution
FUNCTION execute_command(command, args):
active_session = get_active_session_from_registry()
IF active_session EXISTS:
context = load_session_context(active_session.directory)
workspace = active_session.directory
inherit_task_context(context)
ELSE:
context = create_temporary_workspace()
workspace = temporary_directory
execute_with_context(command, args, context, workspace)
END FUNCTION
Protocol Benefits:
- Active Session Discovery: Query
.workflow/session_status.jsonlfor active session - Context Inheritance: Use active session directory and documents for command execution
- Fallback Mode: Commands can operate without active session (creates temporary workspace)
- Output Location: Active session determines where files are created/modified
- Task Context: Active session provides current task purpose and requirements
Individual Session Tracking
All workflow state for each session managed through workflow-session.json with comprehensive structure:
Session State Structure
{
"session_id": "WFS-[topic-slug]",
"session_version": "2.0",
"project": "feature description",
"type": "simple|medium|complex",
"current_phase": "PLAN|IMPLEMENT|REVIEW",
"status": "active|paused|completed",
"created_at": "timestamp",
"updated_at": "timestamp",
"checkpoints": {
"plan": {
"status": "completed|in_progress|pending",
"documents": ["IMPL_PLAN.md", "TASK_BREAKDOWN.md"],
"timestamp": "timestamp"
},
"implement": {
"status": "completed|in_progress|pending",
"agents_completed": ["code-developer"],
"current_agent": "code-review-agent",
"todos": {
"total": 12,
"completed": 8,
"in_progress": 1
},
"timestamp": "timestamp"
},
"review": {
"status": "completed|in_progress|pending",
"quality_checks": {
"code_quality": "passed",
"test_coverage": "pending"
}
}
},
"context_chain": [],
"state_transitions": []
}
Phase-Aware Session Management
Conceptual/Planning Phase
- Tracks planning document generation
- Monitors task decomposition progress
- Preserves planning context and decisions
- Safe interruption at document boundaries
Implementation Phase
- Integrates with existing TodoWrite system
- Tracks agent progression and outputs
- Maintains file modification history
- Supports multi-agent coordination
Review Phase
- Tracks validation and quality gates
- Preserves review comments and decisions
- Maintains compliance check status
Automatic Checkpoints
Checkpoint Triggers
-
Planning Phase:
- After planning document completion
- After task breakdown generation
- On user interrupt request
-
Implementation Phase:
- After agent completion
- At TodoWrite milestones
- After significant file changes
- On phase transitions
-
Review Phase:
- After quality check completion
- On validation milestones
- At review agent boundaries
Checkpoint Strategy
{
"save_triggers": ["agent_complete", "todo_milestone", "user_interrupt"],
"save_data": ["agent_outputs", "file_changes", "todo_state"],
"resume_logic": "skip_completed_continue_sequence"
}
Cross-Phase Context Preservation
Context Chain Maintenance
- All phase outputs preserved in session
- Context automatically transferred between phases
- Planning documents bridge PLAN → IMPLEMENT phases
- Implementation artifacts bridge IMPLEMENT → REVIEW
- Full audit trail maintained for decisions
State Transitions
{
"from": "PLAN",
"to": "IMPLEMENT",
"timestamp": "timestamp",
"trigger": "planning completion",
"handoff_data": {
"plan_path": ".workflow/WFS-[topic-slug]/IMPL_PLAN.md",
"tasks": ["task1", "task2"],
"complexity": "medium"
}
}
Recovery Mechanisms
Automatic Recovery Logic
def resume_workflow():
session = load_session()
if session.current_phase == "PLAN":
resume_planning(session.checkpoints.plan)
elif session.current_phase == "IMPLEMENT":
resume_implementation(session.checkpoints.implement)
elif session.current_phase == "REVIEW":
resume_review(session.checkpoints.review)
State Validation
- Verify required artifacts exist for resumption
- Check file system consistency with session state
- Validate TodoWrite synchronization
- Ensure agent context completeness
- Confirm phase prerequisites met
Recovery Strategies
- Complete Recovery: Full state restoration when possible
- Partial Recovery: Resume with warning when some data missing
- Graceful Degradation: Restart phase with maximum retained context
- Manual Intervention: Request user guidance for complex conflicts
Agent Integration Protocol
Required Agent Capabilities
All agents must support:
- Checkpoint save/load functionality
- State validation for resumption
- Context preservation across interrupts
- Progress reporting to session manager
Phase-Specific Integration
- Planning Agents: Auto-save planning outputs
- Implementation Agents: Track code changes and test results
- Review Agents: Preserve validation outcomes
Error Handling
Common Scenarios
-
Session File Corruption:
- Automatic backup before each save
- Rollback to last known good state
- Recovery from planning documents
-
Version Incompatibility:
- Automatic migration for minor versions
- Manual intervention for major changes
- Backward compatibility for essential fields
-
Missing Dependencies:
- Graceful handling of missing files
- Regeneration of recoverable artifacts
- Clear error messages for resolution
-
Multi-Session Conflicts:
- Registry integrity validation
- Active session collision detection
- Automatic session status correction
Session Lifecycle Management
Complete Session Lifecycle
1. Registration Phase
- Add session to global registry (
.workflow/session_status.jsonl) - Generate unique session ID in WFS-[topic-slug] format
- Create session directory structure
2. Activation Phase
- Set session as active (deactivates any other active session)
- Initialize session state file (
workflow-session.json) - Create base directory structure based on complexity level
3. Execution Phase
- Track progress through workflow phases (PLAN → IMPLEMENT → REVIEW)
- Maintain checkpoints at natural boundaries
- Update session state with phase transitions and progress
4. State Management Phase
- Active: Session is currently being worked on
- Paused: Session temporarily suspended, can be resumed
- Completed: Session finished successfully
5. Session Operations
- Switching: Change active session (preserves state of previous)
- Resumption: Intelligent recovery from saved state and checkpoints
- Interruption: Graceful pause with complete state preservation
Session State Transitions
INACTIVE → ACTIVE → PAUSED → ACTIVE → COMPLETED
↑ ↓ ↓ ↑ ↓
CREATE PAUSE SWITCH RESUME ARCHIVE
Implementation Guidelines
Session Management Operations
Testing Requirements
- Single-phase interruption/resumption
- Multi-phase workflow continuity
- Context preservation validation
- Error recovery scenarios
- Multi-session registry operations
- Session switching without data loss
- Active session inheritance in commands
- Registry integrity validation
- Version migration testing
Success Metrics
- Zero data loss on resume or session switch
- Context continuity maintained across sessions
- No duplicate work performed
- Full workflow completion capability
- Seamless multi-session management
- Registry integrity maintained
- Commands automatically inherit active session context
- Minimal performance overhead