AgentSkillsCN

performance-load-testing

针对负载特征真实性、瓶颈识别与饱和行为的专项工作流程。当必须对吞吐量、延迟或资源效率瓶颈进行度量与缓解时使用;切勿用于非性能相关的功能验收决策。

SKILL.md
--- frontmatter
name: performance-load-testing
description: Specialized workflow for load profile realism, bottleneck detection, and saturation behavior. Use when throughput, latency, or resource-efficiency bottlenecks must be measured and mitigated; do not use for non-performance functional acceptance decisions.

Performance Load Testing

Trigger Boundary

  • Use when latency, throughput, or resource saturation must be measured and improved.
  • Do not use for product roadmap prioritization; use project-estimation or technical-roadmapping.
  • Do not use for compliance control reviews alone; use security-*.

Goal

Improve system performance with measurable and sustainable gains.

Inputs

  • Change scope and risk profile
  • Domain evidence for load profile realism, bottleneck detection, and saturation behavior
  • Operational, compliance, and rollout constraints

Outputs

  • Load test plan with target workload profiles
  • Decision log for load profile realism, bottleneck detection, and saturation behavior
  • Verification checklist with measurable pass-fail criteria

Workflow

  1. Clarify outcomes and hard constraints for load profile realism, bottleneck detection, and saturation behavior.
  2. Produce options and select an approach for load profile realism, bottleneck detection, and saturation behavior.
  3. Evaluate trade-offs across security, performance, operability, and maintainability.
  4. Verify decisions using load-test execution and bottleneck evidence.
  5. Publish decisions, residual risks, and accountable follow-up actions.

Quality Gates

  • Scope and assumptions for load profile realism, bottleneck detection, and saturation behavior are explicit and reviewable.
  • Decision rationale is backed by evidence instead of preference.
  • Rollout and rollback criteria are defined when production impact exists.
  • Residual risks have owners, due dates, and verification steps.

Failure Handling

  • Stop when load test does not cover production-critical traffic patterns.
  • Escalate when accepted risk exceeds team policy thresholds.