- Add task saturation assessment to merge preparation with execution when appropriate - Optimize complexity thresholds: <3/<8 saturated tasks instead of <5/<15 - Enhance JSON schema with preparation_complexity, preparation_tasks, estimated_prep_time - Update execute.md to handle merged tasks with PREPARATION_INCLUDED marker - Add document reference paths for execution phase - Reduce task fragmentation while preserving complex preparation when needed 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
6.7 KiB
Workflow Plan Command (/workflow:plan)
Overview
Creates actionable implementation plans with intelligent input source detection. Supports text, files, issues, and templates through automatic recognition.
Core Principles
File Structure: @~/.claude/workflows/workflow-architecture.md
Usage
/workflow:plan [--AM gemini|codex] <input>
Analysis Method Flag (--AM)
Optional flag to specify which CLI tool to use for context analysis:
- gemini (default): Uses Gemini CLI for pattern-based analysis and architectural understanding
- codex: Uses Codex CLI for autonomous development context gathering with intelligent file discovery
Examples:
/workflow:plan --AM codex "Build authentication system"
/workflow:plan --AM gemini requirements.md
Input Detection Logic
The command automatically detects input type:
File Input (Auto-detected)
/workflow:plan requirements.md
/workflow:plan config.json
Triggers: Extensions: .md, .txt, .json, .yaml, .yml Processing: Reads file contents and extracts requirements
Issue Input (Auto-detected)
/workflow:plan ISS-001
/workflow:plan ISSUE-123
Triggers: Patterns: ISS-, ISSUE-, -request- Processing: Loads issue data and acceptance criteria
Text Input (Default)
/workflow:plan "Build user authentication with JWT and OAuth2"
Triggers: Everything else Processing: Parse natural language requirements
Automatic Behaviors
Session Management
- Creates new session if none exists
- Uses active session if available
- Generates session ID: WFS-[topic-slug]
Complexity Detection with Saturation
- Simple: <3 saturated tasks → Direct IMPL_PLAN.md
- Medium: 3-8 saturated tasks → IMPL_PLAN.md + TODO_LIST.md
- Complex: >8 saturated tasks → Full decomposition
Note: 1 complex preparation task = 0.5 saturated task for counting
Task Generation with Saturation Control
- Task Saturation Assessment: Evaluates whether to merge preparation and execution
- Default merge mode: "Analyze and implement X" instead of "Analyze X" + "Implement X"
- Smart splitting: Only separate when preparation complexity > threshold
- Automatically creates .task/ files when complexity warrants
- Generates hierarchical task structure (max 2 levels)
- Main tasks with subtasks become container tasks (not directly executable)
- Updates session state with task references
- Runs project structure analysis to populate paths field
Task Saturation Assessment
Evaluates whether to merge preparation and execution:
Merge Conditions (Saturated Tasks):
- Preparation work < 30min estimated
- Analysis scope ≤ 3 files
- No complex dependencies
- Single module operation
Separate Preparation Conditions:
- Cross-module architecture analysis (>5 modules)
- Deep performance analysis or security audit required
- Database migration design needed
- Third-party API integration research
Task Examples:
- Saturated: "IMPL-001: Analyze auth patterns and implement JWT token management"
- Split needed: "IMPL-001: Design cross-service auth architecture" + "IMPL-002: Implement auth service"
Task Breakdown Process
- Automatic decomposition: Tasks with complexity >5 are broken into subtasks (impl-N.M format)
- Container tasks: Parent tasks with subtasks become containers (marked with ▸ in TODO_LIST)
- Smart decomposition: AI analyzes task title to suggest logical subtask structure
- Context inheritance: Subtasks inherit parent's requirements and scope, refined for specific needs
- Agent assignment: Automatic agent mapping based on subtask type (planning/code/test/review)
- Maximum depth: 2 levels (impl-N.M) to maintain manageable hierarchy
Implementation Field Requirements
- analysis_source: Determines context gathering method (manual|auto-detected|gemini|codex)
- Auto-assignment: manual (user provides details) → auto-detected (system infers) → gemini/codex (needs analysis)
- Required fields: files (with path/location/modifications), context_notes, analysis_source
- Paths format: Semicolon-separated list (e.g., "src/auth;tests/auth;config/auth.json")
Session Check Process
⚠️ CRITICAL: Check for existing active session before planning
- Check Active Session: Check for
.workflow/.active-*marker file - Session Selection: Use existing active session or create new
- Context Integration: Load session state and existing context
Output Documents
Document References for Execution
Primary References
-
Planning Document:
.workflow/WFS-[topic-slug]/IMPL_PLAN.md- Used by:
/workflow:executefor context loading - Contains: Requirements, task overview, success criteria
- Used by:
-
Task Definitions:
.workflow/WFS-[topic-slug]/.task/impl-*.json- Used by: Agents for implementation context
- Contains: Complete task details with implementation field including preparation_complexity
-
Progress Tracking:
.workflow/WFS-[topic-slug]/TODO_LIST.md- Used by:
/workflow:executefor status tracking - Updated by: Agents after task completion
- Used by:
IMPL_PLAN.md (Always Created)
# Implementation Plan - [Project Name]
*Generated from: [input_source]*
## Requirements
[Extracted requirements from input source]
## Task Breakdown
- **IMPL-001**: [Saturated task description with merged preparation and execution]
- **IMPL-002**: [Task description]
## Success Criteria
[Measurable completion conditions]
Optional TODO_LIST.md (Auto-triggered)
Created when complexity > simple or task count > 5
TODO_LIST Structure: Uses unified hierarchical list format
- Container tasks (with subtasks) marked with
▸symbol - Leaf tasks use standard
- [ ]/- [x]checkboxes - Indentation shows hierarchy (2 spaces per level)
- Container tasks represent logical grouping, not executable items
Task JSON Files (Auto-created)
Generated in .task/ directory when decomposition enabled
Error Handling
Input Processing Errors
- File not found: Clear error message with suggestions
- Invalid issue: Verify issue ID exists
- Unknown template: List available templates
- Empty input: Prompt for valid input
- Vague text input: Auto-reject without guidance
- Rejected examples: "do something", "fix it", "make it better", "add feature"
- Response: Direct rejection message, no further assistance
System ensures: Unified planning interface with intelligent input detection and automatic complexity handling