Request Optimizer

Request Optimizer v2.0

Purpose

Intelligent gateway that intercepts and analyzes every user request, providing strategic recommendations before execution. This skill orchestrates the entire Claude Code ecosystem by:

• Calculating task complexity (0.0 - 1.0)
• Routing to appropriate agents (27 available)
• Activating relevant skills (27 available)
• Coordinating MCPs (14 available)
• Selecting optimal model (Haiku/Sonnet/Opus)
• Preparing for future LimitlessAgent integration

When to Use

This skill should be used automatically on every non-trivial request to:

1. Analyze request specificity and clarity
2. Score complexity using weighted factors
3. Route to appropriate resources
4. Recommend execution strategy
5. Get approval before heavy operations
6. Execute and report results

Architecture

USER REQUEST
     │
     ↓
┌─────────────────────────────────────────────────────────────┐
│                  request-optimizer v2.0                      │
├─────────────────────────────────────────────────────────────┤
│                                                              │
│   PHASE 1: Analysis (5 Points)                              │
│   └─ references/analysis-framework.md                       │
│                                                              │
│   PHASE 2: Complexity Scoring                               │
│   └─ references/complexity-scoring.md                       │
│                                                              │
│   PHASE 3: Resource Selection                               │
│   ├─ references/agent-routing.md (27 agents)                │
│   ├─ references/skill-routing.md (27 skills)                │
│   └─ references/mcp-routing.md (14 MCPs)                    │
│                                                              │
│   PHASE 4: Execution Routing                                │
│   └─ references/decision-tree.md                            │
│                                                              │
│   PHASE 5: Integration (Future)                             │
│   └─ references/integration-interfaces.md                   │
│                                                              │
└─────────────────────────────────────────────────────────────┘
     │
     ↓
┌─────────────────────────────────────────────────────────────┐
│                   EXECUTION PATH                             │
├─────────────────────────────────────────────────────────────┤
│                                                              │
│   Complexity < 0.3   →   Direct Execution (Haiku)           │
│   Complexity 0.3-0.7 →   Agent Execution (Sonnet)           │
│   Complexity > 0.7   →   LimitlessAgent (Opus/NZT)          │
│                                                              │
└─────────────────────────────────────────────────────────────┘

Reference Files

How It Works

Phase 1: Analysis (5 Points)

When a request is received, analyze using references/analysis-framework.md:

1. Specificity Assessment - How specific/vague is the request?
2. Exploration Detection - Does this need codebase exploration?
3. Subtask Identification - Should this be decomposed?
4. Tool Coordination - What resources are needed?
5. Model Recommendation - Which model is optimal?

Phase 2: Complexity Scoring

Calculate complexity using references/complexity-scoring.md:

complexity_score = (
    scope_factor * 0.25 +
    depth_factor * 0.25 +
    ambiguity_factor * 0.20 +
    tooling_factor * 0.15 +
    duration_factor * 0.15
)

Result: 0.0 (trivial) to 1.0 (maximum complexity)

Phase 3: Resource Selection

Based on complexity and request type, consult:

• references/agent-routing.md - Select from 27 specialized agents
• references/skill-routing.md - Select from 27 available skills
• references/mcp-routing.md - Select from 14 MCP integrations

Phase 4: Recommendation

Present structured recommendation:

## Analysis Results

| Factor | Assessment |
|--------|------------|
| **Specificity** | [HIGH/MEDIUM/LOW] |
| **Complexity Score** | [0.0 - 1.0] |
| **Exploration Needed** | [Yes/No] |
| **Subtasks Identified** | [Count] |

## Resource Recommendation

| Type | Resource | Reason |
|------|----------|--------|
| Model | [Haiku/Sonnet/Opus] | [Why] |
| Agent | [name or none] | [Why] |
| Skills | [list or none] | [Why] |
| MCPs | [list or none] | [Why] |

## Execution Path

[Direct | Agent | LimitlessAgent]

## Approval Required

[List of items needing approval]

Ready to execute? (Yes / No / Adjust)

Phase 5: Execution

After approval:

1. Execute recommended strategy
2. Track metrics
3. Report results
4. Ask if additional steps needed

Approval Gates

Always Require Approval

• Invoking Explore Agent
• Invoking any specialized Agent
• Using Opus model
• Escalating to LimitlessAgent
• MCP write operations
• Multi-step workflows (5+ tasks)

Safe Without Approval

• Analyzing request
• Recommending strategy
• Reading files
• Simple edits (1 file, clear scope)
• MCP read operations

External Catalogs

This skill references but does not duplicate:

Future Integration

LimitlessAgent

When complexity > 0.7 and task is multi-step, this skill will:

1. Create IExecutionPlan (see references/integration-interfaces.md)
2. Handoff to LimitlessAgent
3. Monitor execution via NZT Protocol
4. Report results

State Persistence

Future versions will persist:

• Execution history
• Metrics
• Learning patterns

Via Supabase (see references/integration-interfaces.md)

Constraints

1. Always get approval before heavy operations
2. Start with analysis, not execution
3. Be concise in reporting
4. Recommend /clear when context is bloated
5. Default to Haiku for simple tasks
6. Respect rate limits and cost budgets

Changelog

See CHANGELOG.md for version history.