jwcodewrote/agent_skills_plugin

Overview

This skill designs and reviews production systems that must handle high concurrency, high performance, and high availability. It produces concise, actionable architecture plans, SLI/SLO proposals, capacity estimates, and test/operational guides. Use it to align architecture choices with explicit targets for latency, throughput, and resilience.

How this skill works

I gather workload shapes, targets (RPS/QPS, p95/p99 latency, error budget, RTO/RPO) and critical paths, then model request flow and concurrency boundaries. I recommend scaling axes, partitioning keys, caching tiers, and failover strategies, and produce validation plans (load, stress, soak, chaos) and operational artifacts like runbooks and dashboards.

When to use it

• Planning architecture for a new or re-architected production service with strict latency/availability targets
• Designing scaling strategy and capacity plan for expected peak traffic and growth
• Defining SLIs/SLOs, error budgets, and incident readiness for critical systems
• Preparing load, stress, soak, or chaos testing aligned to real SLO targets
• Optimizing data access, caching, and partitioning to reduce tail latency

Best practices

• Collect absolute targets up front: RPS, p95/p99, peak, growth, RTO/RPO, consistency needs
• Choose 3–5 primary SLIs mapped to explicit SLOs; prefer percentiles and availability over averages
• Model concurrency boundaries and bound resources with queues, pools, and backpressure
• Design for graceful degradation and clear failover paths; eliminate single points of failure
• Validate with load, stress, soak, and chaos tests and keep rollback & mitigation plans ready

Example use cases

• Design a horizontally scalable API layer with p99 latency under 200ms and multi-zone availability
• Plan capacity, partitioning and caching for a read-heavy analytics service with bursty traffic
• Define SLIs/SLOs and runbooks for a payments or trading system with strict RTO/RPO
• Create a test plan to validate auto-scaling and failover behavior under simulated outages
• Optimize a microservice’s data access pattern to reduce p99 tail latency using read replicas and materialized views