hardware-performance-expert

name	hardware-performance-expert
description	Expert system for hardware analysis, real-time performance troubleshooting, and LLM optimization.

Skill: Hardware & Performance Expert

This skill is your primary expert for everything related to your machine's physical capabilities and real-time health. It provides deep diagnostics and actionable solutions for performance bottlenecks.

When to use

Activate this skill whenever you need to:

1. Inventory: Get a detailed report of the local hardware (CPU, RAM, GPU, OS).
2. Troubleshoot: Resolve slowness, high resource usage, or thermal throttling.
3. AI Optimization: Evaluate LLM compatibility and optimize inference speeds.
4. Upgrades: Get technical recommendations for hardware improvements.

Performance Troubleshooting Protocol

1. Resource Saturation Analysis

• Tool: analyze_performance, analyze_ram_pressure
• CPU Threshold: If currentLoad > 80% sustained, identify top processes and recommend actions.
• RAM Threshold: If memory_used_percent > 90%, run analyze_ram_pressure to analyze swap usage and top RSS processes.

2. Thermal & Power Bottlenecks

• Context: High-end hardware performing below expectations.
• Action: Call thermal_profile to check for temperature spikes or frequency clipping (throttling) and suggest environmental fixes.

3. Storage Performance

• Tool: analyze_local_system, check_storage_health
• Action: Monitor SSD fill rates (> 90%). Call check_storage_health to verify SMART status and identify legacy HDDs slowing down modern workflows.

4. Dev Environment & Security

• Context: Slow build times or "permission denied" errors on dev paths.
• Action: Call diagnose_antivirus_impact to detect EDR/Antivirus exclusion coverage on project hot paths.

5. Time-Bound Monitoring

• Context: Transient issues that happen during a specific task (e.g., "The machine slows down only when I compile" or "Inference starts fast then lags").
• Action: Call monitor_system_health with a duration (e.g., 30s) to collect statistical data (min/max/avg) while the problematic task is running.

Workflow Integration

1. Initial Scan: Run analyze_local_system + analyze_performance in parallel.
2. Identify Bottleneck: Compare current metrics against theoretical hardware limits.
3. Expert Advice: Provide specific, technical remediation steps.
4. Verification: Confirm the fix with a follow-up performance scan.

Available Tools (Summary)

• analyze_local_system: Hardware inventory.
• analyze_performance: Real-time performance metrics.
• analyze_ram_pressure: Detailed memory pressure and RSS analysis.
• check_storage_health: Disk SMART health and firmware status.
• thermal_profile: Real-time CPU/GPU thermal and frequency state.
• diagnose_antivirus_impact: Antivirus/EDR impact assessment.
• monitor_system_health: Statistical health report over a duration (time-bound monitoring).
• check_llm_compatibility: (BETA) LLM performance prediction (requires remote API).
• get_llm_recommendations: (BETA) AI model advisor (requires remote API).
• analyze_inference_config: Deep-dive for AI runtimes.

Guidelines

• Correlate symptoms with empirical data.
• Provide actionable CLI commands for quick fixes.
• Protect privacy by using anonymized data for remote checks.