Skip to main content

DELIBERATION MODE

Analyze requirements and decompose tasks for: $ARGUMENTS

System Prompt

⚠️ EXECUTION DIRECTIVE: When the user invokes this command, you MUST:

  1. IMMEDIATELY execute - no questions, no explanations first
  2. ALWAYS show full output from script/tool execution
  3. ALWAYS provide summary after execution completes

DO NOT:

  • Say "I don't need to take action" - you ALWAYS execute when invoked
  • Ask for confirmation unless requires_confirmation: true in frontmatter
  • Skip execution even if it seems redundant - run it anyway

The user invoking the command IS the confirmation.


Usage

# Plan a feature implementation
/deliberation "user authentication system"

# Analyze requirements for a task
/deliberation "migrate from REST to GraphQL"

# Decompose complex work
/deliberation "implement real-time collaboration features"

# Clarify ambiguous requirements
/deliberation "optimize database performance"

Mode Rules

❌ STRICTLY FORBIDDEN

  • NO code execution - Analysis only
  • NO tool calls (except Read for context gathering)
  • NO implementation - Planning phase only
  • NO premature solutions - Explore problem space first

✅ REQUIRED ACTIVITIES

  • Pure analysis - Understand requirements deeply
  • Task decomposition - Break into logical phases
  • Gap identification - Find missing information
  • Dependency mapping - Understand relationships
  • Question formulation - Clarify ambiguities

Deliberation Framework

Phase 1: Requirement Analysis

## Requirements Understanding

### Explicit Requirements
1. [Requirement as stated]
2. [Requirement as stated]

### Implied Requirements
1. [Inferred from context]
2. [Inferred from context]

### Ambiguities / Questions
1. [Unclear aspect] - Need clarification on X
2. [Missing detail] - Should we assume Y or Z?

### Constraints
- Technical: [Framework, language, platform]
- Business: [Timeline, resources]
- Dependencies: [External systems]

Phase 2: Task Decomposition

## Task Breakdown

### Phase 1: [Phase Name]
**Duration**: [Estimated time]
**Complexity**: [Low/Medium/High]
**Dependencies**: [What must be done first]

Tasks:
1. [Specific task] - [Why needed]
2. [Specific task] - [Why needed]

### Phase 2: [Phase Name]
[... repeat structure ...]

## Dependency Graph

Phase 1 (Foundation) ├─> Phase 2 (Core) │ └─> Phase 4 (Integration) └─> Phase 3 (Auxiliary) └─> Phase 4 (Integration)

Phase 3: Complexity Assessment

## Complexity Analysis

### Token Budget Estimate
- Phase 1: 15K tokens (research + design)
- Phase 2: 25K tokens (implementation)
- Phase 3: 10K tokens (testing)
- **Total**: 50K tokens

### Risk Zones
- **High Risk**: [Aspect with high complexity]
- **Medium Risk**: [Aspect with moderate complexity]
- **Low Risk**: [Well-understood aspects]

### Recommended Strategy
Based on complexity: [Use /complexity-gauge, /recursive-workflow, orchestrator, etc.]

Phase 4: Alternative Approaches

## Approaches Considered

### Approach 1: [Name]
**Pros**: [Benefits]
**Cons**: [Drawbacks]
**Complexity**: [Assessment]
**Recommendation**: [Yes/No/Maybe]

### Approach 2: [Name]
[... repeat structure ...]

### Recommended Approach
[Chosen approach with rationale]

Output Structure

# Deliberation: [Feature/Task Name]

## Executive Summary
[2-3 sentences: What we're building, why, high-level approach]

## Requirement Analysis
[See Phase 1 above]

## Task Decomposition
[See Phase 2 above]

## Complexity Assessment
[See Phase 3 above]

## Alternative Approaches
[See Phase 4 above]

## Questions for User
1. [Clarifying question 1]
2. [Clarifying question 2]

## Next Steps
**If approved, transition to**: RESEARCH mode (verify assumptions) or ACTION mode (implementation)

**Ready to proceed?**

Integration

  • Auto-load: multi-agent-workflow skill (token estimation)
  • Auto-load: framework-patterns skill (architecture options)
  • Use: /complexity-gauge for token budget analysis

Transition Protocol

After deliberation complete:

**Deliberation complete. Ready to transition:**

Option 1: RESEARCH mode
- Verify technical assumptions
- Check if packages/APIs exist
- Validate approach feasibility
- Estimated: 5-10 tool calls

Option 2: ACTION mode (if confident)
- Skip verification
- Proceed directly to implementation
- Use deliberation plan as blueprint

Which mode should we use next?

Best Practices

DO:

  • Ask clarifying questions
  • Consider multiple approaches
  • Identify dependencies explicitly
  • Estimate complexity honestly
  • Note assumptions clearly

DON'T:

  • Jump to implementation
  • Make unjustified assumptions
  • Skip dependency analysis
  • Ignore edge cases
  • Provide only one approach

Action Policy

<default_behavior> This command analyzes and recommends without making changes. Provides:

  • Detailed analysis of current state
  • Specific recommendations with justification
  • Prioritized action items
  • Risk assessment

User decides which recommendations to implement. </default_behavior>

After analysis, provide: - Analysis completeness (all aspects covered) - Recommendation confidence levels - Specific examples from codebase - Clear next steps for user

Success Output

When deliberation completes:

✅ COMMAND COMPLETE: /deliberation
Topic: <feature/task>
Requirements: N explicit, N implied
Phases: N identified
Complexity: <Low|Medium|High>
Approaches: N considered
Questions: N for user

Completion Checklist

Before marking complete:

  • Requirements analyzed
  • Tasks decomposed
  • Complexity assessed
  • Approaches considered
  • Questions formulated

Failure Indicators

This command has FAILED if:

  • ❌ No requirements identified
  • ❌ No task breakdown
  • ❌ No complexity estimate
  • ❌ No next steps

When NOT to Use

Do NOT use when:

  • Simple task (no planning needed)
  • Already have clear plan
  • Implementation phase started

Anti-Patterns (Avoid)

Anti-PatternProblemSolution
Jump to implementationMissing requirementsComplete deliberation first
Single approachMissed alternativesConsider multiple options
Skip questionsWrong assumptionsAsk clarifying questions

Principles

This command embodies:

  • #5 No Assumptions - Ask clarifying questions
  • #9 Based on Facts - Analyze requirements
  • #6 Clear, Understandable - Structured output

Full Standard: CODITECT-STANDARD-AUTOMATION.md