Skip to content
Back to skills

Load Test

ASecurity

WAY 15 — API load / stress / spike / soak testing with k6. Use when the user says "load test", "stress test", "perf test the api", "how many users can it handle", "spike test".

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 12, 2026
testingtestingapiperformance

Works with

  • api

Security analysis

A100/100

Scanned October 12, 2026

npx -y skills add GhassenAZZOUZ/claude-qa-agent --skill load-test --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Load Test?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Load Test
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/ghassenazzouz-load-test/badge)](https://www.skillsdirectory.com/skills/ghassenazzouz-load-test)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: load-test
description: WAY 15 — API load / stress / spike / soak testing with k6. Use when the user says "load test", "stress test", "perf test the api", "how many users can it handle", "spike test".
---

# WAY 15 — Load Testing

Run CLAUDE.md **Step 0** first. **Refuse `--env prod`.** Load tests degrade
the system they hit. Point the user to staging or a dedicated perf environment.

## Step 1: Gather inputs
- OpenAPI spec (file or URL) or an endpoint list (required)
- Profile: `smoke` (default) | `load` | `stress` | `spike` | `soak`, plus optional target VUs and duration
- Write endpoints (POST/PUT/PATCH/DELETE): excluded unless the user lists them explicitly **and** confirms the environment isn't shared

## Phase 1: Write + smoke
Run `load-test-writer` with the profile and `run: false` for anything heavier than smoke. It writes the k6 script(s) to `./qa-artifacts/load-tests/<api-slug>-<YYYY-MM-DD>/` and always runs the **smoke** profile, to prove the script works and get baseline latency.

## Phase 2: Confirm, then run the real profile
Show the user the profile shape (VUs × duration), the endpoints, and the target URL. Ask for explicit confirmation. Then run `load-test-writer` again with `run: true`. Don't skip this step for `stress`, `spike` or `soak`.

## Phase 3: Report
1. Per-endpoint table: p95, p99, error rate, threshold ✅/❌, and smoke vs profile latency
2. For each threshold breach, apply the `bug-report` skill (API context, label `performance`). Include the profile, VUs, the p95 vs threshold, and the error rate. Ask the user before filing, because a breach in an under-provisioned staging environment may not be a product bug
3. Save `./qa-artifacts/sessions/load-<api-slug>-<YYYY-MM-DD-HH-MM>.md` with the watermark footer

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.