Claude-Code-Workflow/.claude/workflows/workflow-architecture.md

Workflow Architecture

Overview & Core Principles

This document defines a streamlined workflow system architecture featuring JSON-only data model, marker-based session management, and unified file structure with dynamic task decomposition.

Design Philosophy

• JSON-first data model - JSON files are single source of truth; markdown is generated views
• Marker-based sessions - Ultra-simple active session tracking via marker files
• Unified structure - Same file template for all workflows, created on-demand
• Dynamic decomposition - Tasks break down as needed during execution
• On-demand creation - Files and directories created only when required
• Agent autonomy - Complete context preserved for independent execution

Data Architecture

JSON-Only Data Model

Core Principle: JSON files are the authoritative source; markdown documents are read-only generated views.

• Task State: Stored exclusively in .task/IMPL-*.json files
• Documents: Generated on-demand from JSON data
• No Synchronization: Eliminates bidirectional sync complexity
• Performance: Direct JSON access without parsing overhead

Task JSON Schema

All task files use this simplified 5-field schema:

{
  "id": "IMPL-1.2",
  "title": "Implement JWT authentication",
  "status": "pending|active|completed|blocked|container",

  "meta": {
    "type": "feature|bugfix|refactor|test|docs",
    "agent": "code-developer|planning-agent|code-review-test-agent"
  },

  "context": {
    "requirements": ["JWT authentication", "OAuth2 support"],
    "focus_paths": ["src/auth", "tests/auth", "config/auth.json"],
    "acceptance": ["JWT validation works", "OAuth flow complete"],
    "parent": "IMPL-1",
    "depends_on": ["IMPL-1.1"],
    "inherited": {
      "from": "IMPL-1",
      "context": ["Authentication system design completed"]
    },
    "shared_context": {
      "auth_strategy": "JWT with refresh tokens"
    }
  },

  "flow_control": {
    "pre_analysis": [
      {
        "step": "gather_context",
        "action": "Read dependency summaries",
        "command": "bash(cat .workflow/*/summaries/IMPL-1.1-summary.md)",
        "output_to": "auth_design_context",
        "on_error": "skip_optional"
      }
    ],
    "implementation_approach": {
      "task_description": "Implement comprehensive JWT authentication system...",
      "modification_points": ["Add JWT token generation...", "..."],
      "logic_flow": ["User login request → validate credentials...", "..."]
    },
    "target_files": [
      "src/auth/login.ts:handleLogin:75-120",
      "src/middleware/auth.ts:validateToken"
    ]
  }
}

Focus Paths Field Details

The focus_paths field within context specifies concrete project paths relevant to the task implementation:

Focus Paths Format

• Array of strings: ["folder1", "folder2", "specific_file.ts"]
• Concrete paths: Use actual directory/file names without wildcards
• Mixed types: Can include both directories and specific files
• Relative paths: From project root (e.g., src/auth, not ./src/auth)

Path Selection Strategy

• Directories: Include relevant module directories (e.g., src/auth, tests/auth)
• Specific files: Include files explicitly mentioned in requirements (e.g., config/auth.json)
• Avoid wildcards: Use concrete paths discovered via get_modules_by_depth.sh
• Focus scope: Only include paths directly related to task implementation

Examples

// Authentication system task
"focus_paths": ["src/auth", "tests/auth", "config/auth.json", "src/middleware/auth.ts"]

// UI component task
"focus_paths": ["src/components/Button", "src/styles", "tests/components"]

// Database migration task
"focus_paths": ["migrations", "src/models", "config/database.json"]

Flow Control Field Details

The flow_control field serves as a universal process manager for task execution with comprehensive flow orchestration:

pre_analysis Array - Sequential Process Steps

Each step contains:

• step: Unique identifier for the step
• action: Human-readable description of what the step does
• command: Executable command wrapped in bash() with embedded context variables (e.g., bash(command with [variable_name]))
• output_to: Variable name to store step results (optional for final steps)
• on_error: Error handling strategy (skip_optional, fail, retry_once, manual_intervention)
• success_criteria: Optional validation criteria (e.g., exit_code:0)

Context Flow Management

• Variable Accumulation: Each step can reference outputs from previous steps via [variable_name]
• Context Inheritance: Steps can use dependency summaries and parent task context
• Pipeline Processing: Results flow sequentially through the analysis chain

Variable Reference Format

• Context Variables: Use [variable_name] to reference step outputs
• Task Properties: Use [depends_on], [focus_paths] to reference task JSON properties
• Bash Compatibility: Avoids conflicts with bash ${} variable expansion

Command Types Supported

• CLI Analysis: bash(~/.claude/scripts/gemini-wrapper -p 'prompt')
• Agent Execution: bash(codex --full-auto exec 'task description')
• Shell Commands: bash(cat), bash(grep), bash(find), bash(custom scripts)
• Context Processing: bash(file reading), bash(dependency loading), bash(context merging)

Error Handling Strategies

• skip_optional: Continue execution, step result is empty
• fail: Stop execution, mark task as failed
• retry_once: Retry step once, then fail if still unsuccessful
• manual_intervention: Pause execution for manual review

Example Flow Control Patterns

Pattern 1: Multi-Step Analysis

"pre_analysis": [
  {
    "step": "gather_dependencies",
    "command": "bash(for dep in ${depends_on}; do cat .summaries/$dep-summary.md; done)",
    "output_to": "dependency_context"
  },
  {
    "step": "analyze_codebase",
    "command": "bash(gemini -p '@{[focus_paths]} using context: [dependency_context]')",
    "output_to": "codebase_analysis"
  }
]

Pattern 2: Direct Implementation

"pre_analysis": [
  {
    "step": "implement",
    "command": "bash(codex --full-auto exec 'Implement [title] in [focus_paths]')"
  }
]

Benefits of Flow Control

• Universal Process Manager: Handles any type of analysis or implementation flow
• Context Accumulation: Builds comprehensive context through step chain
• Error Recovery: Granular error handling at step level
• Command Flexibility: Supports any executable command or agent
• Dependency Integration: Automatic loading of prerequisite task results

Task Hierarchy

Structure: Maximum 2 levels (IMPL-N.M format)

IMPL-1              # Main task
IMPL-1.1            # Subtask (dynamically created)
IMPL-1.2            # Another subtask
IMPL-2              # Another main task
IMPL-2.1            # Subtask (dynamically created)

Rules:

• Container tasks: Parent tasks with subtasks (cannot execute directly)
• Leaf tasks: Only executable tasks
• Status inheritance: Parent status derived from subtask completion

Session Management

Active Session Marker System

Ultra-Simple Tracking: .workflow/.active-[session-name]

.workflow/
├── WFS-oauth-integration/         # Session directory (paused)
├── WFS-user-profile/             # Session directory (paused)
├── WFS-bug-fix-123/              # Session directory (completed)
└── .active-WFS-user-profile      # Marker file (active session)

Benefits:

• Zero Parsing: File existence check is atomic
• Atomic Operations: File creation/deletion is naturally atomic
• Visual Discovery: ls .workflow/.active-* shows active session
• Simple Switching: Delete old + create new marker

Session Operations

# Detect active session
active_session=$(find .workflow -name ".active-*" | head -1)
if [ -n "$active_session" ]; then
  session_name=$(basename "$active_session" | sed 's/^\.active-//')
  echo "Active session: $session_name"
fi

# Switch session
find .workflow -name ".active-*" -delete && touch .workflow/.active-WFS-new-feature

Session State

Each session directory contains workflow-session.json:

{
  "session_id": "WFS-[topic-slug]",
  "project": "feature description",
  "type": "simple|medium|complex",
  "current_phase": "PLAN|IMPLEMENT|REVIEW",
  "status": "active|paused|completed",
  "progress": {
    "completed_phases": ["PLAN"],
    "current_tasks": ["IMPL-1", "IMPL-2"]
  }
}

File Organization

Unified Structure

All workflows use the same structure regardless of complexity. Files created on-demand only.

.workflow/WFS-[topic-slug]/
├── workflow-session.json        # Session state (REQUIRED)
├── IMPL_PLAN.md                # Planning document (REQUIRED)
├── TODO_LIST.md                # Progress tracking (REQUIRED)
├── .task/                      # Task definitions (REQUIRED)
│   ├── IMPL-*.json             # Main task definitions
│   └── IMPL-*.*.json           # Subtask definitions (dynamic)
├── [.brainstorming/]           # Brainstorming phase (on-demand)
├── [.chat/]                    # CLI sessions (on-demand)
│   ├── chat-*.md              # Saved chat sessions
│   └── analysis-*.md          # Analysis results
└── [.summaries/]               # Task summaries (on-demand)
    ├── IMPL-*.md              # Main task summaries
    └── IMPL-*.*.md            # Subtask summaries

Creation Strategy

• Initial: Only workflow-session.json, IMPL_PLAN.md, TODO_LIST.md, .task/
• On-Demand:
  • .brainstorming/ → When brainstorming initiated
  • .chat/ → When analysis commands executed
  • .summaries/ → When first task completed

Naming Conventions

• Sessions: WFS-[topic-slug] (lowercase with hyphens)
• Conflicts: Add -NNN suffix if needed
• Documents: Standard names (workflow-session.json, IMPL_PLAN.md, etc.)
• Chat: chat-analysis-*.md
• Summaries: IMPL-[task-id]-summary.md

Workflow Classification

Complexity Types

Based on task count and decomposition needs:

Type	Tasks	Depth	Behavior
Simple	≤5	1 level	Direct execution, minimal decomposition
Medium	6-10	2 levels	Moderate decomposition, context coordination
Over-scope	>10	Re-scope	Requires project re-scoping into iterations

Characteristics

Simple: Bug fixes, small features, config changes

• Single-level tasks, direct execution, ≤5 tasks

Medium: New features, API endpoints, schema changes

• Two-level hierarchy, context coordination, 6-10 tasks

Over-scope: Major features exceeding 10 tasks

• Requires re-scoping into smaller iterations, never >10 tasks

Assessment Rules

• Creation: System evaluates and assigns complexity
• 10-task limit: Hard limit enforced - exceeding requires re-scoping
• Execution: Can upgrade (Simple→Medium→Over-scope), triggers re-scoping
• Override: Users can manually specify complexity within 10-task limit

Task Execution Framework

Agent Integration

Assignment Rules (based on task type/keywords):

• Planning → planning-agent
• Implementation → code-developer
• Testing → code-review-test-agent
• Review → review-agent

Execution Context (agents receive):

{
  "task": { /* complete task JSON */ },
  "workflow": {
    "session": "WFS-user-auth",
    "phase": "IMPLEMENT"
  }
}

Operations Guide

Basic Operations

# Session initialization
mkdir -p .workflow/WFS-topic-slug/.task
echo '{"session_id":"WFS-topic-slug",...}' > .workflow/WFS-topic-slug/workflow-session.json
echo '# Implementation Plan' > .workflow/WFS-topic-slug/IMPL_PLAN.md
echo '# Tasks' > .workflow/WFS-topic-slug/TODO_LIST.md

# Task creation
echo '{"id":"IMPL-1","title":"New task",...}' > .task/IMPL-1.json

# Task updates
jq '.status = "active"' .task/IMPL-1.json > temp && mv temp .task/IMPL-1.json

# Document generation
generate_todo_list_from_json .task/

Validation Rules

1. ID Uniqueness: All task IDs must be unique
2. Format: IMPL-N[.M] pattern (max 2 levels)
3. References: Parent IDs must exist as JSON files
4. Fields: All 5 core fields required (id, title, status, meta, context, flow_control)
5. Paths: context.focus_paths must contain valid project paths
6. Dependencies: context.depends_on task IDs must exist

Error Recovery

• Missing Session: Remove orphaned active marker
• Multiple Active: Keep newest, remove others
• Corrupted Session: Recreate from template
• Broken Hierarchy: Reconstruct parent-child relationships

# Session consistency check
active_marker=$(find .workflow -name ".active-*" | head -1)
if [ -n "$active_marker" ]; then
  session_name=$(basename "$active_marker" | sed 's/^\.active-//')
  session_dir=".workflow/$session_name"
  if [ ! -d "$session_dir" ]; then
    echo "⚠️ Orphaned active marker, removing..."
    rm "$active_marker"
  fi
fi

Reference Templates

Integration Points

• [Component Name]: [how to use/integrate]
• [API Endpoint]: [usage details]
• [Configuration]: [location and format]

Testing Verification

• [Validation]: [confirmation method]
• [Quality checks]: [what was verified]

Notes for Future Tasks

[Any important considerations, limitations, or follow-up items]

### TODO List (TODO_LIST.md)
```markdown
# Tasks: [Session Topic]

## Task Progress
▸ **IMPL-001**: [Main Task Group] → [📋](./.task/IMPL-001.json)
  - [ ] **IMPL-001.1**: [Subtask] → [📋](./.task/IMPL-001.1.json)
  - [x] **IMPL-001.2**: [Subtask] → [📋](./.task/IMPL-001.2.json) | [✅](./.summaries/IMPL-001.2.md)

- [x] **IMPL-002**: [Simple Task] → [📋](./.task/IMPL-002.json) | [✅](./.summaries/IMPL-002.md)

▸ **IMPL-003**: [Main Task Group] → [📋](./.task/IMPL-003.json)
  - [ ] **IMPL-003.1**: [Subtask] → [📋](./.task/IMPL-003.1.json)
  - [ ] **IMPL-003.2**: [Subtask] → [📋](./.task/IMPL-003.2.json)

## Status Legend
- `▸` = Container task (has subtasks)
- `- [ ]` = Pending leaf task
- `- [x]` = Completed leaf task
- Maximum 2 levels: Main tasks and subtasks only

## Notes
[Optional notes]

---

System Design: Unified workflow architecture featuring ultra-fast session management, JSON-only data model, and consistent file structure across all workflow types.