Agent Skills: Thinking Framework v4.0 - Multi-Agent Systematic Problem-Solving

Thinking Framework v4.0 - Multi-Agent Systematic Problem-Solving

Purpose: Decompose complex problems and derive optimal solutions using structured thinking methods with multi-agent orchestration.

When to Use

• Complex problem-solving requiring systematic decomposition
• Root cause analysis (finding the "why" behind issues)
• Strategic planning (strengths/weaknesses, competitive analysis)
• Decision-making under uncertainty
• Innovation requiring creative breakthroughs

Multi-Agent Architecture (v4.0)

Agent Tiers

| Tier | Agent | Role | Clear-Thought Tools | |------|-------|------|---------------------| | 1 | Orchestrator | Workflow coordination, complexity routing | decisionframework, metacognitivemonitoring | | 2 | ProblemDefiner | Problem clarification, decomposition | sequentialthinking, mentalmodel | | 2 | MethodExecutor | Thinking method execution | Method-specific (see mapping) | | 2 | StrategyArchitect | Strategic synthesis, action planning | collaborativereasoning, decisionframework |

Complexity-Based Routing

| Complexity | Indicators | Agent Configuration | Time | |------------|-----------|---------------------|------| | Simple | Single cause, 1-2 steps, clear path | Orchestrator only | <30s | | Medium | 3-5 factors, some ambiguity | + 1 Specialist (sequential) | 30-60s | | Complex | 5+ factors, high interdependencies | + 2-3 Specialists (parallel) | 60s+ |

Clear-Thought Tool Mapping

| Method | Primary Tool | Secondary Tool | |--------|--------------|----------------| | 5 Why | sequentialthinking | - | | Fishbone | collaborativereasoning | visualreasoning | | First Principles | mentalmodel | sequentialthinking | | SWOT | decisionframework | - | | OODA Loop | scientificmethod | sequentialthinking | | Dialectic | structuredargumentation | - | | Design Thinking | collaborativereasoning | mentalmodel | | Pareto | decisionframework | mentalmodel | | PDCA | scientificmethod | - | | GAP Analysis | visualreasoning | decisionframework | | Kepner-Tregoe | decisionframework | structuredargumentation | | TRIZ | mentalmodel | designpattern | | SCAMPER | collaborativereasoning | - | | DMAIC | scientificmethod | metacognitivemonitoring |

Execution Routines

A. Divide & Conquer (Complex Only)

When: Systemic problems with 5+ interdependent factors

Agent Flow:

Orchestrator → ProblemDefiner → MethodExecutor(s) [parallel] → StrategyArchitect → Orchestrator

Process:

1. Orchestrator: Assess complexity, dispatch ProblemDefiner
2. ProblemDefiner: Define problem, decompose into ≤5 sub-problems
3. MethodExecutor(s): Analyze sub-problems in parallel
   • Each executor uses appropriate Clear-Thought tool
4. StrategyArchitect: Synthesize findings, create action plan
5. Orchestrator: Quality gate, final integration

Output:

## Problem Definition
[Clear statement from ProblemDefiner]

## Sub-Problem Analyses
| Sub-Problem | Method | Root Cause | Recommendation |
|-------------|--------|------------|----------------|
| SP1 | [method] | [cause] | [action] |

## Integrated Strategy
[From StrategyArchitect]

## Quality Assessment
- Confidence: [%]
- Uncertainties: [list]

## Core Insight
[One sentence]

B. Method Selection (All Cases)

When: Any problem, especially Simple-Medium complexity

Agent Flow:

• Simple: Orchestrator only (direct method application)
• Medium: Orchestrator → MethodExecutor → Orchestrator

Process:

1. Classify problem type
2. Select method using matching matrix
3. Execute with appropriate Clear-Thought tool
4. Output optimized format

Method-Problem Matching:

| Problem Type | Methods | Clear-Thought Tool | |--------------|---------|-------------------| | root_cause | 5 Why, Fishbone | sequentialthinking, collaborativereasoning | | creative_innovation | SCAMPER, TRIZ, Design Thinking | collaborativereasoning, mentalmodel | | strategic_planning | SWOT + 2x2, GAP Analysis | decisionframework, visualreasoning | | process_improvement | Pareto, PDCA, GAP | decisionframework, scientificmethod | | decision_making | OODA Loop, Kepner-Tregoe | scientificmethod, decisionframework |

C. Strategy Routine (Strategic Decisions)

When: Strategic planning with strengths/weaknesses analysis

Agent Flow:

Orchestrator → ProblemDefiner → MethodExecutor (SWOT) → StrategyArchitect → Orchestrator

Process:

1. Diagnose: Strengths (with evidence) + Weaknesses (root cause via 5 Why)
2. Analyze: Use decisionframework for SWOT evaluation
3. Strategize: StrategyArchitect creates 2x2 matrix
4. Plan: GAP Analysis → Action items

2x2 Matrix (MANDATORY):

           │ Maximize Strengths │ Address Weaknesses │
───────────┼────────────────────┼────────────────────┤
High       │   DO FIRST         │   REMOVE RISK      │
Priority   │   (Invest now)     │   (Critical fix)   │
───────────┼────────────────────┼────────────────────┤
Low        │   LONG-TERM R&D    │   STRATEGIC IGNORE │
Priority   │   (Future bet)     │   (Accept risk)    │

Core Strategy Template:

"Maximize [strength] through [method], address [weakness] via [action], to achieve [goal]."

Quality Gates

| Gate | Stage | Check | Tool | |------|-------|-------|------| | G1 | Problem Definition | Clarity, specificity, boundedness | metacognitivemonitoring | | G2 | Method Selection | Problem-method fit | decisionframework | | G3 | Analysis | Depth, evidence, logic | metacognitivemonitoring | | G4 | Integration | Coherence, completeness, actionability | metacognitivemonitoring |

Gate Protocol:

• Complex: All gates mandatory
• Medium: G2 + G4
• Simple: G4 only

Output Guidelines

Format Selection:

• Structured comparisons → Markdown tables
• Sequential processes → Numbered lists
• Problem decomposition → Mermaid diagrams
• Strategic decisions → 2x2 Matrix

Always Include:

• Complexity assessment (pre-flight)
• Method selection justification
• Confidence score
• One-sentence summary

Quick Reference

• Agent Prompts: agents/
• Method Catalog: reference/INDEX.md
• 60-sec Selector: reference/QUICK_SELECTOR.md
• Multi-Method Workflows: reference/METHOD_COMBINATIONS.md
• Agent Patterns: reference/AGENT_PATTERNS.md
• Practical Examples: GUIDE.md

Anti-Patterns

| Pattern | Problem | Solution | |---------|---------|----------| | Over-engineering | A Routine for simple problems | Use complexity assessment | | Under-analysis | Simple method for complex problems | Proper routing | | Tool mismatch | Wrong Clear-Thought tool for method | Follow mapping table | | Skip quality gates | Missing validation | Enforce gate protocol | | Sequential when parallel | Slow complex analysis | Use parallel agents |

Meta

After analysis, briefly reflect:

• What worked? What could improve?
• Was method optimal? Was tool mapping effective?
• Agent coordination smooth?