What it does
Builds an executable K6 performance script for an endpoint or journey, encodes the response-time/error-rate/throughput SLAs from qa.config.yml as run-failing thresholds, and optionally smoke-runs it against a non-prod target. Tests ISO/IEC 25010 performance efficiency (and reliability for soak).
When to use it
After /qa:perf-plan defines the workload model and SLAs, to execute load, stress, spike, or soak testing.
Prerequisites
qa.config.yml with tooling.performance: k6 and gates.performance SLAs; a non-prod target if running; ideally a PERF-PLAN-*.md for the workload model.
Output
ISO/IEC/IEEE 29119-3 Test Execution Log (K6 JSON summary) under paths.reports_dir, plus the K6 Test Procedure under paths.tests_dir/performance/.
Mechanics
How it works
- Read config; gate that tooling.performance is k6
- Resolve target/type from arguments (default load; default target to risk_areas.critical)
- Derive workload model from PERF-PLAN-*.md or gates.performance
- Build the K6 script with per-type stages (load/stress/spike/soak)
- Encode gates.performance SLAs as options.thresholds so a breach fails the run
- Wire the runner and JSON summary export
- Optionally smoke-run against non-prod and report p95/p99/error-rate vs thresholds
Why it works
The theory behind it
Performance testing execution (CT-PT Specialist). Thresholds-as-gates make performance efficiency (and soak reliability) objectively pass/fail, not subjective.
CT-PT (Specialist) · ISO/IEC 25010 performance efficiency / reliability · CTFL v4.0 §2.2.2
Example
See it in use
> /qa:perf-test "/checkout" load✓ Correct when Thresholds equal the gates (p95 <= 500ms, p99 <= 1000ms, error <= 1%, 100 rps) and the build fails if breached.