Wreck Load

Page speed and load testing for a local web app with the wreck-it CLI. It measures each page with Lighthouse on a simulated phone (LCP, CLS, TBT), diagnoses slow pages with Chrome DevTools performance traces when available, runs ramp, spike and soak load profiles (p50/p90/p99 latency, error rate, breaking point, memory trend), then traces slow endpoints to causes such as N+1 queries, missing indexes or blocking work. Use as stage 7 of a wreck-it run, or when the user asks to "stress test my API", "load test", "how many users can my app handle", "find slow endpoints", "check page speed", or "ru

Testing & QA Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add miyannishar/wreck-it

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add miyannishar/wreck-it --skill wreck-load
Manual install
git clone --depth 1 https://github.com/miyannishar/wreck-it.git /tmp/wreck-it
mkdir -p ~/.claude/skills
cp -r /tmp/wreck-it/skills/wreck-load ~/.claude/skills/wreck-load

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# wreck-load

`WRECK` means `npx @miyanishar/wreck-it`.

`WRECK run stage load running`

## Safety

- GET/HEAD only. Mutating load needs `"allowMutatingLoad": true`, which only the user sets.
- A hosted (remote) database: ask before any load test; load on a production database hurts real users.
- The CLI refuses endpoints that call AI, email, SMS or payments (thousands of real calls) unless the user allowed them.
- Prefer a production build (`npm run build && npm start`) when the user agrees; dev servers give misleading numbers. Say which you tested.

## 0. Page speed (Lighthouse)

```bash
WRECK perf --record                 # discovered pages on a simulated mid-range phone, slow 4G
WRECK perf / /products --device desktop --record   # optional: key pages on desktop
```

- Each page gets a Lighthouse score plus **LCP** (largest paint), **CLS** (layout shift) and **TBT** (main-thread blocking); reports go to `.wreck-it/perf/*.html`. With `--record`, metrics in Google's "poor" range become `performance` findings when a second run agrees.
- On a **dev server** only CLS is recorded (LCP/TBT there reflect unminified code): tell the user and offer a production-build re-run.
- Its "to judge yourself" list (best practices, SEO): mention notable ones; record only what affects users (e.g. a page without a `<title>`).
Read the full skill on GitHub

Details

Skill name
wreck-load
Path
skills/wreck-load/SKILL.md
License
MIT
Repo updated
25 October 2026

FAQ

What does the Wreck Load skill do?

Page speed and load testing for a local web app with the wreck-it CLI. It measures each page with Lighthouse on a simulated phone (LCP, CLS, TBT), diagnoses slow pages with Chrome DevTools performance traces when available, runs ramp, spike and soak load profiles (p50/p90/p99 latency, error rate, breaking point, memory trend), then traces slow endpoints to causes such as N+1 queries, missing indexes or blocking work. Use as stage 7 of a wreck-it run, or when the user asks to "stress test my API", "load test", "how many users can my app handle", "find slow endpoints", "check page speed", or "ru

How do I install the Wreck Load skill?

Claude Code plugin: /plugin marketplace add miyannishar/wreck-it. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Related skills

Wreck Chaos

miyannishar/wreck-it

Chaos-test a running local web app as the chaos-monkey persona, using a page sweep, an API fuzzer (plus Schemathesis when there's an OpenAPI spec), hostile form inputs, double-submits, back/forward/refresh mid-flow, multiple tabs, offline/slow/500-injected network, expired or tampered sessions, an axe-core scan and keyboard-only navigation. Records reproducible bugs with wreck-it. Use as stage 6 of a wreck-it run, or when the user asks to "break my forms", "chaos test", "try weird inputs", "test offline", or "check accessibility".

Testing & QAIn 1 list

Wreck Explore

miyannishar/wreck-it

Explore a running local web app through Playwright MCP as a normal user (every core flow plus data-integrity checks) and then in character as each persona from .wreck-it/personas.md, recording reproducible bugs with wreck-it. Use as stages 4–5 of a wreck-it run, or when the user asks to "click through my app like a user", "test the happy paths", "try my app as a beginner", or "check my core flows".

Testing & QAIn 1 list

Wreck It

miyannishar/wreck-it

Full wreck-it run against a locally running web app. It tests like normal users, ICP personas, rushed beginners, chaos monkeys and traffic spikes, then writes a scored bug report with source file:line and regression Playwright tests. Use when the user says "wreck my app", "wreck it", "test my app like a user", "find bugs before launch", "is my app ready to ship", or asks for a readiness score. Also for one feature, such as "test the checkout", "just test signup", "check the coupon flow" (a focused run). For a single stage ("just stress test my API", "only check accessibility"), use the matchin

Testing & QAIn 1 list

Wreck Personas

miyannishar/wreck-it

Derive realistic user personas (the app's ideal customers plus built-in normal-user, rushed-beginner and chaos-monkey) and their missions for wreck-it testing, saved to .wreck-it/personas.md. Use as stage 3 of a wreck-it run, or when the user asks "who would use my app", "make test personas", or wants to edit the personas used for testing.

Testing & QAIn 1 list

Smoke

danybgoode/golden-frijoles

Verifies real, rendered behavior on this project's frontend (local/preview/staging/prod, unauthed or authed) via a real headless-Chromium Playwright browser — the DEFAULT way to check "does this render correctly / does this look right / is this actually live" for any coding agent, not just Claude Code. Use when asked to "smoke test", "verify the live/prod page", "check if X actually rendered", "run a browser smoke", "does this look right on preview", "verify an authed role's flow", or as a post-merge/pre-PR verification step. Wraps the project's scripts/live-smoke.mjs — a real screenshot + JSO

Testing & QAIn 1 list

Prove It Ran

agentkitco/agentkit-dev-skills

Verify that tests, builds and checks actually ran and exercised the changed code, before claiming a task is done. Use before reporting completion, or when asked to prove tests ran or verify a fix.

Testing & QAIn 1 list

Browse the most popular Claude skills, all testing & qa skills, or write your own with the skills generator.