This ASE skill is built around ghz, the open source gRPC benchmarking and load testing tool maintained in the bojand/ghz project. The agent job is concrete: point ghz at a gRPC method using a proto file, a protoset bundle, or server reflection, replay request data under controlled concurrency or rate schedules, and return structured latency, throughput, and error results. That makes the entry skill-shaped instead of a generic CLI listing, because the useful outcome is not “install ghz.” The useful outcome is “exercise a specific gRPC endpoint with known request fixtures and compare the service behavior under load.”

Invoke this when an agent needs to validate a performance-sensitive gRPC change, reproduce a latency regression, smoke-test a service contract after deployment, or benchmark a method before tuning infrastructure. ghz supports unary and streaming calls, JSON and binary payloads, config files, metadata templates, and multiple output formats including summary, JSON, CSV, HTML, and Influx-friendly formats. That makes it practical for CI checks, staging environment probes, incident follow-up, and repeatable engineering baselines.

The scope boundary is clear. This skill is not a general observability platform, not an HTTP load tester, and not an always-on monitoring system. It is a synthetic gRPC benchmarking workflow that depends on explicit service definitions or reflection and reproducible request fixtures. Integration points include protobuf repos, generated protosets, reflection-enabled services, JSON or TOML test configs, CI pipelines, and downstream reporting tools. If an agent needs to answer “how does this gRPC method behave under a defined load pattern?” ghz is a strong upstream fit.