mirror of https://github.com/catlog22/Claude-Code-Workflow.git synced 2026-02-12 02:37:45 +08:00

Files

catlog22 0380cbb7b8 feat(lite-execute, lite-plan): store original user input for enhanced task execution context

2025-11-17 16:34:23 +08:00

27 KiB

Raw Blame History

name, description, argument-hint, allowed-tools, timeout, color

name	description	argument-hint	allowed-tools	timeout	color
lite-plan	Lightweight interactive planning workflow with in-memory planning, code exploration, and execution dispatch to lite-execute after user confirmation	[-e\|--explore] "task description"\|file.md	TodoWrite(), Task(), SlashCommand(), AskUserQuestion()	180000	cyan

Workflow Lite-Plan Command (/workflow:lite-plan)

Overview

Intelligent lightweight planning command with dynamic workflow adaptation based on task complexity. Focuses on planning phases (exploration, clarification, planning, confirmation) and delegates execution to /workflow:lite-execute.

Core Functionality

Intelligent Task Analysis: Automatically determines if exploration/planning agents are needed
Dynamic Exploration: Calls cli-explore-agent only when task requires codebase understanding
Interactive Clarification: Asks follow-up questions after exploration to gather missing information
Adaptive Planning:
- Simple tasks: Direct planning by current Claude
- Complex tasks: Delegates to cli-planning-agent for detailed breakdown
Two-Step Confirmation: First display complete plan as text, then collect three-dimensional input (task approval + execution method + code review tool)
Execution Dispatch: Stores execution context in memory and calls /workflow:lite-execute --in-memory for actual implementation

Usage

Command Syntax

/workflow:lite-plan [FLAGS] <TASK_DESCRIPTION>

# Flags
-e, --explore              Force code exploration phase (overrides auto-detection logic)

# Arguments
<task-description>         Task description or path to .md file (required)

Execution Process

Workflow Overview

User Input ("/workflow:lite-plan \"task\"")
    |
    v
[Phase 1] Task Analysis & Exploration Decision (10-20 seconds)
    -> Analyze task description
    -> Decision: Need exploration? (Yes/No)
    -> If Yes: Launch cli-explore-agent
    -> Output: exploration findings (if performed)
    |
    v
[Phase 2] Clarification (Optional, user interaction)
    -> If exploration revealed ambiguities or missing info
    -> AskUserQuestion: Gather clarifications
    -> Update task context with user responses
    -> If no clarification needed: Skip to Phase 3
    |
    v
[Phase 3] Complexity Assessment & Planning (20-60 seconds)
    -> Assess task complexity (Low/Medium/High)
    -> Decision: Planning strategy
       - Low: Direct planning (current Claude)
       - Medium/High: Delegate to cli-planning-agent
    -> Output: Task breakdown with execution approach
    |
    v
[Phase 4] Task Confirmation & Execution Selection (User interaction)
    -> Step 4.1: Output complete plan as text to user
    -> Step 4.2: AskUserQuestion with three dimensions
       1. Confirm task: Allow/Modify/Cancel (multi-select, can supplement via Other)
       2. Execution method: Agent/Codex/Auto (single-select, auto: simple→agent, complex→codex)
       3. Code review: Skip/Gemini/Agent/Other (single-select, can specify custom tool via Other)
    -> Process selections and proceed to Phase 5
    -> If cancel: Exit
    |
    v
[Phase 5] Dispatch to Execution
    -> Store execution context (plan, exploration, clarifications, selections)
    -> Call /workflow:lite-execute --in-memory
    -> Execution delegated to lite-execute command
    |
    v
Planning Complete (Execution continues in lite-execute)

Task Management Pattern

lite-plan focuses on planning phases (1-4) with TodoWrite tracking
Phase 5 stores execution context and dispatches to lite-execute
Execution tracking (TodoWrite updates, progress monitoring) handled by lite-execute
No intermediate file artifacts generated (all planning in-memory)
Execution artifacts (code changes, test results) managed by lite-execute

Detailed Phase Execution

Phase 1: Task Analysis & Exploration Decision

Operations:

Analyze task description to determine if code exploration is needed

Decision logic:

needsExploration = (
  flags.includes('--explore') || flags.includes('-e') ||  // Force exploration if flag present
  (
    task.mentions_specific_files ||
    task.requires_codebase_context ||
    task.needs_architecture_understanding ||
    task.modifies_existing_code
  )
)

Decision Criteria:

Task Type	Needs Exploration	Reason
Any task with `-e` or `--explore` flag	Yes (forced)	Flag overrides auto-detection logic
"Implement new feature X"	Maybe	Depends on integration with existing code
"Refactor module Y"	Yes	Needs understanding of current implementation
"Add tests for Z"	Yes	Needs to understand code structure
"Create new standalone utility"	No	Self-contained, no existing code context
"Update documentation"	No	Doesn't require code exploration
"Fix bug in function F"	Yes	Needs to understand implementation

If Exploration Needed:

Launch cli-explore-agent with task-specific focus

Agent call format:

Task(
  subagent_type="cli-explore-agent",
  description="Analyze codebase for task context",
  prompt=`
  Task: ${task_description}

  Analyze and return the following information in structured format:
  1. Project Structure: Overall architecture and module organization
  2. Relevant Files: List of files that will be affected by this task (with paths)
  3. Current Implementation Patterns: Existing code patterns, conventions, and styles
  4. Dependencies: External dependencies and internal module dependencies
  5. Integration Points: Where this task connects with existing code
  6. Architecture Constraints: Technical limitations or requirements
  7. Clarification Needs: Ambiguities or missing information requiring user input

  Time Limit: 60 seconds

  Output Format: Return a JSON-like structured object with the above fields populated.
  Include specific file paths, pattern examples, and clear questions for clarifications.
  `
)

Expected Return Structure:

explorationContext = {
  project_structure: "Description of overall architecture",
  relevant_files: ["src/auth/service.ts", "src/middleware/auth.ts", ...],
  patterns: "Description of existing patterns (e.g., 'Uses dependency injection pattern', 'React hooks convention')",
  dependencies: "List of dependencies and integration points",
  integration_points: "Where this connects with existing code",
  constraints: "Technical constraints (e.g., 'Must use existing auth library', 'No breaking changes')",
  clarification_needs: [
    {
      question: "Which authentication method to use?",
      context: "Found both JWT and Session patterns",
      options: ["JWT tokens", "Session-based", "Hybrid approach"]
    },
    // ... more clarification questions
  ]
}

Output Processing:

Store exploration findings in explorationContext
Extract clarification_needs array from exploration results
Set needsClarification = (clarification_needs.length > 0)
Use clarification_needs to generate Phase 2 questions

Progress Tracking:

Mark Phase 1 as completed
If needsClarification: Mark Phase 2 as in_progress
Else: Skip to Phase 3

Expected Duration: 10-20 seconds (analysis) + 30-60 seconds (exploration if needed)

Phase 2: Clarification (Optional)

Skip Condition: Only run if Phase 1 set needsClarification = true

Operations:

Review explorationContext.clarification_needs from Phase 1
Generate AskUserQuestion based on exploration findings
Focus on ambiguities that affect implementation approach

AskUserQuestion Call (simplified reference):

// Use clarification_needs from exploration to build questions
AskUserQuestion({
  questions: explorationContext.clarification_needs.map(need => ({
    question: `${need.context}\n\n${need.question}`,
    header: "Clarification",
    multiSelect: false,
    options: need.options.map(opt => ({
      label: opt,
      description: `Use ${opt} approach`
    }))
  }))
})

Output Processing:

Collect user responses and store in clarificationContext
Format: { question_id: selected_answer, ... }
This context will be passed to Phase 3 planning

Progress Tracking:

Mark Phase 2 as completed
Mark Phase 3 as in_progress

Expected Duration: User-dependent (typically 30-60 seconds)

Phase 3: Complexity Assessment & Planning

Operations:

Assess task complexity based on multiple factors
Select appropriate planning strategy
Generate task breakdown using selected method

Complexity Assessment Factors:

complexityScore = {
  file_count: exploration.files_to_modify.length,
  integration_points: exploration.dependencies.length,
  architecture_changes: exploration.requires_architecture_change,
  technology_stack: exploration.unfamiliar_technologies.length,
  task_scope: (task.estimated_steps > 5),
  cross_cutting_concerns: exploration.affects_multiple_modules
}

// Calculate complexity
if (complexityScore < 3) complexity = "Low"
else if (complexityScore < 6) complexity = "Medium"
else complexity = "High"

Complexity Levels:

Level	Characteristics	Planning Strategy
Low	1-2 files, simple changes, clear requirements	Direct planning (current Claude)
Medium	3-5 files, moderate integration, some ambiguity	Delegate to cli-planning-agent
High	6+ files, complex architecture, high uncertainty	Delegate to cli-planning-agent with detailed analysis

Planning Execution:

Option A: Direct Planning (Low Complexity)

Current Claude generates plan directly following these guidelines:

Summary: 2-3 sentence overview of the implementation
Approach: High-level implementation strategy
Task Breakdown: 3-5 specific, actionable tasks with file paths
Estimated Time: Total implementation time estimate
Recommended Execution: "Agent" (for Low complexity tasks)

Option B: Agent-Based Planning (Medium/High Complexity)

Delegate to cli-planning-agent with detailed requirements:

Task(
  subagent_type="cli-planning-agent",
  description="Generate detailed implementation plan",
  prompt=`
  Task: ${task_description}
  Exploration Context: ${JSON.stringify(explorationContext, null, 2)}
  User Clarifications: ${JSON.stringify(clarificationContext, null, 2) || "None provided"}
  Complexity Level: ${complexity}

  Generate a detailed implementation plan with the following components:

  1. Summary: 2-3 sentence overview of the implementation
  2. Approach: High-level implementation strategy
  3. Task Breakdown: 3-10 specific, actionable tasks
     - Each task should specify: What to do, Which files to modify/create
  4. Estimated Time: Total implementation time estimate
  5. Recommended Execution: "Agent" or "Codex" based on task complexity

  Ensure tasks are specific, with file paths and clear acceptance criteria.
  `
)

Expected Return Structure (Both Options):

planObject = {
  summary: string,              // 2-3 sentence overview
  approach: string,             // High-level implementation strategy
  tasks: string[],              // 3-10 tasks with file paths
  estimated_time: string,       // Total implementation time estimate
  recommended_execution: string, // "Agent" (Low) or "Codex" (Medium/High)
  complexity: string            // "Low" | "Medium" | "High"
}

Progress Tracking:

Mark Phase 3 as completed
Mark Phase 4 as in_progress

Expected Duration:

Low complexity: 20-30 seconds (direct)
Medium/High complexity: 40-60 seconds (agent-based)

Phase 4: Task Confirmation & Execution Selection

User Interaction Flow: Two-step confirmation process

Step 4.1: Display Plan Summary

First, output the complete plan to the user as regular text:

## Implementation Plan

**Summary**: ${planObject.summary}

**Approach**: ${planObject.approach}

**Task Breakdown**:
${planObject.tasks.map((t, i) => `${i+1}. ${t}`).join('\n')}

**Complexity**: ${planObject.complexity}
**Estimated Time**: ${planObject.estimated_time}
**Recommended Execution**: ${planObject.recommended_execution}

Step 4.2: Collect User Confirmation

After displaying the plan, collect three inputs via AskUserQuestion:

Operations:

Collect four inputs:
1. Task confirmation (multi-select: Allow/Modify/Cancel + optional supplements via "Other")
2. Execution method (single-select: Agent/Codex/Auto)
  - Agent: Execute with @code-developer
  - Codex: Execute with codex CLI tool
  - Auto: Simple tasks (Low complexity) → Agent, Complex tasks (Medium/High) → Codex
3. Code review tool (single-select: Skip/Gemini/Agent + custom tools via "Other")
  - Gemini Review: Use gemini CLI for code analysis
  - Agent Review: Use @code-reviewer agent
  - Other: Specify custom tool (e.g., "qwen", "codex") via text input
4. Task JSON output (single-select: Yes/No)
  - Yes: Export plan to task JSON file for reuse or documentation
  - No: Keep plan in-memory only
Support plan supplements and custom tool specification via "Other" input

Four Questions in Single AskUserQuestion Call:

Question 1: Task confirmation (multi-select: Allow/Modify/Cancel)
Question 2: Execution method selection (single-select: Agent/Codex/Auto)
Question 3: Code review tool selection (single-select: Skip/Gemini/Agent, custom via "Other")
Question 4: Task JSON output (single-select: Yes/No)

AskUserQuestion Call:

AskUserQuestion({
  questions: [
    {
      question: `**Plan Summary**: ${planObject.summary}

**Tasks**: ${planObject.tasks.length} tasks | **Complexity**: ${planObject.complexity} | **Estimated Time**: ${planObject.estimated_time}

Confirm this plan? (Multi-select enabled - you can select multiple options and add supplements via "Other")`,
      header: "Confirm Plan",
      multiSelect: true,
      options: [
        { label: "Allow", description: "Proceed with plan as-is" },
        { label: "Modify", description: "Adjust plan before execution" },
        { label: "Cancel", description: "Abort workflow" }
      ]
    },
    {
      question: `Select execution method:`,
      header: "Execution",
      multiSelect: false,
      options: [
        { label: "Agent", description: "Execute with @code-developer agent" },
        { label: "Codex", description: "Execute with codex CLI tool" },
        { label: "Auto", description: `Auto-select: ${planObject.complexity === 'Low' ? 'Agent (Low complexity)' : 'Codex (Medium/High complexity)'}` }
      ]
    },
    {
      question: `Enable code review after execution?

(You can specify other tools like "qwen" or "codex" via "Other" option)`,
      header: "Code Review",
      multiSelect: false,
      options: [
        { label: "Gemini Review", description: "Review with Gemini CLI tool (gemini-2.5-pro)" },
        { label: "Agent Review", description: "Review with @code-reviewer agent" },
        { label: "Skip", description: "No review needed" }
      ]
    },
    {
      question: `Export plan to task JSON file?

This allows you to reuse the plan or use it with lite-execute later.`,
      header: "Export JSON",
      multiSelect: false,
      options: [
        { label: "Yes", description: "Export plan to JSON file (recommended for complex tasks)" },
        { label: "No", description: "Keep plan in-memory only" }
      ]
    }
  ]
})

Decision Flow:

Task Confirmation (Multi-select):
  ├─ Allow (+ optional supplements in Other) → Proceed to Execution Method Selection
  ├─ Modify (+ optional supplements in Other) → Re-run Phase 3 with modifications
  └─ Cancel → Exit (no execution)

Execution Method Selection (Single-select):
  ├─ Agent → Launch @code-developer agent
  ├─ Codex → Execute with codex CLI tool
  └─ Auto → Automatic selection:
      ├─ If complexity = Low → Launch @code-developer agent
      └─ If complexity = Medium/High → Execute with codex CLI tool

Code Review Selection (after execution):
  ├─ Skip → Skip review, workflow complete
  ├─ Gemini Review → Run gemini code analysis (gemini-2.5-pro)
  ├─ Agent Review → Current Claude agent review
  └─ Other → Specify custom tool (e.g., "qwen", "codex") via text input

Task JSON Export:
  ├─ Yes → Export plan to JSON file before execution
  │        Format: .workflow/lite-plans/plan-{timestamp}.json
  │        Contains: planObject + explorationContext + clarificationContext
  └─ No → Keep plan in-memory only, no file export

Progress Tracking:

Mark Phase 4 as completed
Mark Phase 5 as in_progress

Expected Duration: User-dependent (1-3 minutes typical)

Phase 5: Dispatch to Execution

Operations:

Export plan to task JSON file (if user selected "Yes")
Store execution context in memory variable
Call lite-execute command with --in-memory flag
Execution tracking delegated to lite-execute

Step 5.1: Export Task JSON (Optional)

Skip Condition: Only run if user selected "Yes" for Task JSON Export in Phase 4

Operations:

Create .workflow/lite-plans/ directory if not exists
Generate timestamp-based filename
Export complete plan context to JSON file

// Only execute if userSelection.export_task_json === "Yes"
if (userSelection.export_task_json === "Yes") {
  const timestamp = new Date().toISOString().replace(/[:.]/g, '-')
  const filename = `.workflow/lite-plans/plan-${timestamp}.json`

  const planExport = {
    metadata: {
      created_at: new Date().toISOString(),
      task_description: original_task_description,
      complexity: planObject.complexity,
      estimated_time: planObject.estimated_time
    },
    plan: planObject,
    exploration: explorationContext || null,
    clarifications: clarificationContext || null
  }

  Write(filename, JSON.stringify(planExport, null, 2))

  // Display export confirmation to user
  console.log(`Plan exported to: ${filename}`)
  console.log(`You can reuse this plan with: /workflow:lite-execute ${filename}`)
}

Export File Format:

{
  "metadata": {
    "created_at": "2025-01-17T10:30:00.000Z",
    "task_description": "Original task description",
    "complexity": "Medium",
    "estimated_time": "30 minutes"
  },
  "plan": {
    "summary": "...",
    "approach": "...",
    "tasks": [...],
    "recommended_execution": "Agent|Codex",
    "complexity": "Low|Medium|High"
  },
  "exploration": {...} or null,
  "clarifications": {...} or null
}

Step 5.2: Store Execution Context

Create execution context variable with all necessary information:

// Create execution context for lite-execute
executionContext = {
  planObject: planObject,              // From Phase 3
  explorationContext: explorationContext || null,  // From Phase 1
  clarificationContext: clarificationContext || null, // From Phase 2
  executionMethod: userSelection.execution_method,    // From Phase 4
  codeReviewTool: userSelection.code_review_tool,    // From Phase 4
  originalUserInput: original_task_description        // Original user input
}

Context Structure:

{
  planObject: {
    summary: string,
    approach: string,
    tasks: string[],
    estimated_time: string,
    recommended_execution: string,
    complexity: string
  },
  explorationContext: {...} | null,
  clarificationContext: {...} | null,
  executionMethod: "Agent" | "Codex" | "Auto",
  codeReviewTool: "Skip" | "Gemini Review" | "Agent Review" | string,
  originalUserInput: string  // User's original task description
}

Step 5.3: Call lite-execute

Dispatch execution to lite-execute command:

SlashCommand(command="/workflow:lite-execute --in-memory")

Execution Handoff:

lite-execute reads executionContext variable
All execution logic (TodoWrite tracking, Agent/Codex calls, code review) handled by lite-execute
lite-plan completes after successful handoff

Progress Tracking:

Mark Phase 5 as completed
Execution tracking handed over to lite-execute
User sees execution progress from lite-execute TodoWrite updates

Expected Duration: < 1 second (dispatch only, actual execution in lite-execute)

Best Practices

Workflow Intelligence

Dynamic Adaptation: Workflow automatically adjusts based on task characteristics
- Smart exploration: Only runs when task requires codebase context
- Adaptive planning: Simple tasks get direct planning, complex tasks use specialized agent
- Context-aware clarification: Only asks questions when truly needed
- Reduces unnecessary steps while maintaining thoroughness
- Flag-Based Control:
  - Use -e or --explore to force exploration when:
    - Task appears simple but you know it requires codebase context
    - Auto-detection might miss subtle integration points
    - You want comprehensive code understanding before planning
Progressive Clarification: Gather information at the right time
- Phase 1: Explore codebase to understand current state
- Phase 2: Ask clarifying questions based on exploration findings
- Phase 3: Plan with complete context (task + exploration + clarifications)
- Avoids premature assumptions and reduces rework
Complexity-Aware Planning: Planning strategy matches task complexity
- Low complexity (1-2 files): Direct planning by current Claude (fast, 20-30s)
- Medium complexity (3-5 files): CLI planning agent (detailed, 40-50s)
- High complexity (6+ files): CLI planning agent with risk analysis (thorough, 50-60s)
- Balances speed and thoroughness appropriately
Two-Step Confirmation Process: Clear plan presentation followed by comprehensive control
- Step 1: Display complete plan as readable text output (not embedded in question)
  - Shows summary, approach, tasks, dependencies, risks, complexity, time estimate
  - Clear separation between plan content and user interaction
- Step 2: Collect three-dimensional input via AskUserQuestion
  - First dimension: Confirm/Modify/Cancel plan (multi-select with supplement via "Other")
  - Second dimension: Execution method selection (Agent/Codex/Auto)
  - Third dimension: Code review tool selection (Skip/Gemini/Agent, custom via "Other")
- Allows plan refinement without re-selecting execution method
- Supports iterative planning with user feedback
- Auto mode intelligently selects execution method based on complexity
- Custom code review tools (qwen, codex, etc.) can be specified via "Other" option

Task Management

Live Progress Tracking: TodoWrite provides real-time execution call visibility
- Execution calls ([Agent-1], [Codex-1], etc.) created before execution starts
- Updated in real-time as execution calls progress
- User sees current execution call being worked on (e.g., "[Agent-1] (Task A + Task B)")
- Each execution call shows task summary for context
- Clear completion status at call level (not individual task level)
Phase-Based Organization: 5 distinct phases with clear transitions
- Phase 1: Task Analysis & Exploration (automatic)
- Phase 2: Clarification (conditional, interactive)
- Phase 3: Planning (automatic, adaptive)
- Phase 4: Confirmation (interactive, two-dimensional)
- Phase 5: Execution & Tracking (automatic with live updates)
Flexible Task Counts: Task breakdown adapts to complexity
- Low complexity: 3-5 tasks (focused)
- Medium complexity: 5-7 tasks (detailed)
- High complexity: 7-10 tasks (comprehensive)
- Avoids artificial constraints while maintaining focus
Dependency Tracking: Medium/High complexity tasks include dependencies
- Explicit task ordering when sequence matters
- Parallel execution hints when tasks are independent
- Risk flagging for complex interactions
- Helps agent/CLI execute correctly

Planning Standards

Context-Rich Planning: Plans include all relevant context
- Exploration findings (code structure, patterns, constraints)
- User clarifications (requirements, preferences, decisions)
- Complexity assessment (risks, dependencies, time estimates)
- Execution recommendations (Direct vs CLI, specific tool)
Modification Support: Plans can be iteratively refined
- User can request plan modifications in Phase 4
- Feedback incorporated into re-planning
- No need to restart from scratch
- Supports collaborative planning workflow
No File Artifacts: All planning stays in memory
- Faster workflow without I/O overhead
- Cleaner workspace
- Plan context passed directly to execution
- Reduces complexity and maintenance

Error Handling

Common Errors

Error	Cause	Resolution
Phase 1 Exploration Failure	cli-explore-agent unavailable or timeout	Skip exploration, set `explorationContext = null`, log warning, continue to Phase 2/3 with task description only
Phase 2 Clarification Timeout	User no response > 5 minutes	Use exploration findings as-is without clarification, proceed to Phase 3 with warning
Phase 3 Planning Agent Failure	cli-planning-agent unavailable or timeout	Fallback to direct planning by current Claude (simplified plan), continue to Phase 4
Phase 3 Planning Timeout	Planning takes > 90 seconds	Generate simplified direct plan, mark as "Quick Plan", continue to Phase 4 with reduced detail
Phase 4 Confirmation Timeout	User no response > 5 minutes	Save plan context to temporary var, display resume instructions, exit gracefully
Phase 4 Modification Loop	User requests modify > 3 times	Suggest breaking task into smaller pieces or using /workflow:plan for comprehensive planning

Input/Output

Input Requirements

Task description: String or path to .md file (required)
- Should be specific and concrete
- Can include context about existing code or requirements
- Examples:
  - "Implement user authentication with JWT tokens"
  - "Refactor logging module for better performance"
  - "Add unit tests for authentication service"
Flags (optional):
- -e or --explore: Force code exploration phase (overrides auto-detection)

Output Format

In-Memory Plan Object:

{
  summary: "2-3 sentence overview of implementation",
  approach: "High-level implementation strategy",
  tasks: [
    "Task 1: Specific action with file locations",
    "Task 2: Specific action with file locations",
    // ... 3-7 tasks total
  ],
  complexity: "Low|Medium|High",
  recommended_execution: "Agent|Codex",  // Based on complexity
  estimated_time: "X minutes"
}

Execution Result:

Immediate dispatch to selected tool/agent with plan context
No file artifacts generated during planning phase
Execution starts immediately after user confirmation
Tool/agent handles implementation and any necessary file operations

27 KiB Raw Blame History

Workflow Lite-Plan Command (/workflow:lite-plan)

Overview

Core Functionality

Usage

Command Syntax

Execution Process

Workflow Overview

Task Management Pattern

Detailed Phase Execution

Phase 1: Task Analysis & Exploration Decision

Phase 2: Clarification (Optional)

Phase 3: Complexity Assessment & Planning

Phase 4: Task Confirmation & Execution Selection

Phase 5: Dispatch to Execution

Best Practices

Workflow Intelligence

Task Management

Planning Standards

Error Handling

Common Errors

Input/Output

Input Requirements

Output Format

27 KiB

Raw Blame History