domains / agent-tooling / multi-agent-orchestration-optimization / agent-orchestration-multi-agent-optimize

agent-orchestration-multi-agent-optimize

Skill card

combined

The sickn33 copy is a hardened, safer variant of the rmyndharis copy. It shares the same core structure and instructions but adds critical execution boundaries (disjoint write ownership, cooperative cancellation, context compression validation) and explicitly reframes the code examples as non-executable design sketches. It also corrects the cost optimization section to use verified provider catalogs rather than hardcoded prices. The rmyndharis copy is the original, less constrained version. Because the sickn33 copy contains material safety and implementation constraints that the rmyndharis copy lacks, and the rmyndharis copy presents the code as executable examples (a conflict in interpretation), crowning one would hide the other's specific approach to safety and code presentation.

Repositories

Other copies

Combined skill

---
name: agent-orchestration-multi-agent-optimize
description: "Optimize multi-agent systems with coordinated profiling, workload distribution, and cost-aware orchestration. Use when improving agent performance, throughput, or reliability."
---

# Multi-Agent Optimization Toolkit

## Use this skill when

- Improving multi-agent coordination, throughput, or latency
- Profiling agent workflows to identify bottlenecks
- Designing orchestration strategies for complex workflows
- Optimizing cost, context usage, or tool efficiency

## Do not use this skill when

- You only need to tune a single agent prompt
- There are no measurable metrics or evaluation data
- The task is unrelated to multi-agent orchestration

## Instructions

1. Establish baseline metrics and target performance goals.
2. Profile agent workloads and identify coordination bottlenecks.
3. Apply orchestration changes and cost controls incrementally.
4. Validate improvements with repeatable tests and rollbacks.

## Execution boundaries

This guide does not authorize spawning agents, remote execution, deployment or external messages. Where parallel work is authorized, assign disjoint write ownership, cap concurrency and deadlines, isolate failures, inspect every result and verify the aggregate. A running future cannot generally be stopped by cancelling its handle; design cooperative cancellation and record incomplete work. Context compression must retain constraints, evidence provenance and unresolved tasks; validate losses against a full-context baseline.

## Safety

- Avoid deploying orchestration changes without regression testing.
- Roll out changes gradually to prevent system-wide regressions.

## Role: AI-Powered Multi-Agent Performance Engineering Specialist

### Context

This is an orchestration design guide, not a bundled framework. Profile the existing system and compare scoped changes against a measured baseline. Named agent classes and tracker helpers in design examples must be implemented in the application.

### Core Capabilities

- Intelligent multi-agent coordination
- Performance profiling and bottleneck identification
- Adaptive optimization strategies
- Cross-domain performance optimization
- Cost and efficiency tracking

## Arguments Handling

The tool processes optimization arguments with flexible input parameters:

- `$TARGET`: Primary system/application to optimize
- `$PERFORMANCE_GOALS`: Specific performance metrics and objectives
- `$OPTIMIZATION_SCOPE`: Depth of optimization (quick-win, comprehensive)
- `$BUDGET_CONSTRAINTS`: Cost and resource limitations
- `$QUALITY_METRICS`: Performance quality thresholds

## 1. Multi-Agent Performance Profiling

### Profiling Strategy

- Distributed performance monitoring across system layers
- Real-time metrics collection and analysis
- Continuous performance signature tracking

#### Profiling Agents

1. **Database Performance Agent**
   - Query execution time analysis
   - Index utilization tracking
   - Resource consumption monitoring

2. **Application Performance Agent**
   - CPU and memory profiling
   - Algorithmic complexity assessment
   - Concurrency and async operation analysis

3. **Frontend Performance Agent**
   - Rendering performance metrics
   - Network request optimization
   - Core Web Vitals monitoring

### Profiling design sketch (not executable)

```text
def multi_agent_profiler(target_system):
    agents = [
        DatabasePerformanceAgent(target_system),
        ApplicationPerformanceAgent(target_system),
        FrontendPerformanceAgent(target_system)
    ]

    performance_profile = {}
    for agent in agents:
        performance_profile[agent.__class__.__name__] = agent.profile()

    return aggregate_performance_metrics(performance_profile)
```

## 2. Context Window Optimization

### Optimization Techniques

- Intelligent context compression
- Semantic relevance filtering
- Dynamic context window resizing
- Token budget management

### Context compression design sketch (not executable)

```text
def compress_context(context, max_tokens=4000):
    # Semantic compression using embedding-based truncation
    compressed_context = semantic_truncate(
        context,
        max_tokens=max_tokens,
        importance_threshold=0.7
    )
    return compressed_context
```

## 3. Agent Coordination Efficiency

### Coordination Principles

- Parallel execution design
- Minimal inter-agent communication overhead
- Dynamic workload distribution
- Fault-tolerant agent interactions

### Orchestration design sketch (not executable)

```text
class MultiAgentOrchestrator:
    def __init__(self, agents):
        self.agents = agents
        self.execution_queue = PriorityQueue()
        self.performance_tracker = PerformanceTracker()

    def optimize(self, target_system):
        # Parallel agent execution with coordinated optimization
        with concurrent.futures.ThreadPoolExecutor() as executor:
            futures = {
                executor.submit(agent.optimize, target_system): agent
                for agent in self.agents
            }

            for future in concurrent.futures.as_completed(futures):
                agent = futures[future]
                result = future.result()
                self.performance_tracker.log(agent, result)
```

## 4. Parallel Execution Optimization

### Key Strategies

- Asynchronous agent processing
- Workload partitioning
- Dynamic resource allocation
- Minimal blocking operations

## 5. Cost Optimization Strategies

### LLM Cost Management

- Token usage tracking
- Adaptive model selection
- Caching and result reuse
- Efficient prompt engineering

### Cost decision procedure

Use the application's verified current provider catalog and measured evaluation results, not hardcoded model prices. For each candidate, record input/output unit prices, cache/tool charges, estimated tokens and quality/latency results. Exclude candidates below the predefined quality thresholds.

## 6. Latency Reduction Techniques

### Performance Acceleration

- Predictive caching
- Pre-warming agent contexts
- Intelligent result memoization
- Reduced round-trip communication

## 7. Quality vs Speed Tradeoffs

### Optimization Spectrum

- Performance thresholds
- Acceptable degradation margins
- Quality-aware optimization
- Intelligent compromise selection

## 8. Monitoring and Continuous Improvement

### Observability

- Continuous performance monitoring
- Alerting on performance degradation
- Regular performance reviews