# Verifier Setup

Set a repo up to prove engineering-task work actually works before it ships.

- Repository: https://github.com/AI-Builder-Club/skills (1308 stars)
- Category: Testing
- License: see repository
- Last updated: 2026-09-15
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add AI-Builder-Club/skills --skill verifier-setup -g -a claude-code -y`
Codex: `npx skills add AI-Builder-Club/skills --skill verifier-setup -g -a codex -y`
Cursor: `npx skills add AI-Builder-Club/skills --skill verifier-setup -g -a cursor -y`
Gemini CLI: `npx skills add AI-Builder-Club/skills --skill verifier-setup -g -a gemini-cli -y`
Claude Code plugin marketplace: `/plugin marketplace add AI-Builder-Club/skills`

## Description

Set a repo up to prove engineering-task work actually works before it ships. Investigates the repo, ensures a one-command dev stack (`dev-local`) exists, asks whether verification runs locally or in a sandbox (crabbox), confirms/installs the driver (the `playwright-cli` skill for web by default). Outputs three artifacts: a committed `/verify` skill (per-task verification SOP, spawn a verifier sub-agent → drive the app → screenshot/video proof → open a PR with the proof embedded), the `/dev-local` skill + script, and the installed driver skill. Use when someone says "set up verification", "make this repo verifiable", "scaffold a verify skill", "set up the verifier".

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# verifier-setup, scaffold this repo's `/verify` skill

Goal: leave the repo able to *prove an engineering task works before it ships*, run once, and it wires up everything the per-task `/verify` loop needs.

You are **setting up**, not verifying anything yourself right now. The `/verify`
template lives at `assets/verify.template.md` (next to this skill). Parallels
`dev-local-setup` (which generates a script + its skill doc): a setup skill that
leaves behind reusable, repo-specific artifacts.

## What this produces (the outputs)
Running `verifier-setup` end-to-end leaves the repo with:
1. **A `/verify` skill**, `.claude/skills/verify/SKILL.md`, the repo-tailored
   per-task verification SOP (spawn a verifier sub-agent → drive the app →
   screenshot/video proof → open a PR with the proof embedded). Generated in Step 5.
2. **A `/dev-local` skill + its script**, `scripts/dev-local.sh` **and**
   `.claude/skills/dev-local/SKILL.md`, via `dev-local-setup` (Step 2) if not
   already present. The one-command stack `/verify` depends on.
3. **The driver skill installed**, the `playwright-cli` skill for web apps (Step 2);
   for non-web, the concrete exercise tool confirmed present.

## Step 0, Inventory what already exists (check before you add ANYTHING)
Before creating anything, take stock, the repo may already have some of this,
under whatever name or layout its team chose. Look for the **capability**, not a
specific filename; the paths below are only examples. For each, decide **reuse
as-is / adapt-extend / create fresh**, never blindly overwrite working setup:

The rest: https://github.com/AI-Builder-Club/skills/blob/HEAD/skills/verifier-setup/SKILL.md