Agent Skills › crbnos/carbon › smoke-test

smoke-test

GitHub

对本地ERP开发服务器进行端到端冒烟测试，验证核心模块页面能否正常加载渲染。适用于启动服务后或重大变更后的快速健康检查，不深入业务逻辑。

.claude/skills/smoke-test/SKILL.md crbnos/carbon

Trigger Scenarios

启动开发环境栈后进行广泛范围的代码变更后用户明确要求执行'冒烟测试'时

Install

npx skills add crbnos/carbon --skill smoke-test -g -y

More Options

Non-standard path

npx skills add https://github.com/crbnos/carbon/tree/main/.claude/skills/smoke-test -g -y

Use without installing

npx skills use crbnos/carbon@smoke-test

指定 Agent (Claude Code)

npx skills add crbnos/carbon --skill smoke-test -a claude-code -g -y

安装 repo 全部 skill

npx skills add crbnos/carbon --all -g -y

预览 repo 内 skill

npx skills add crbnos/carbon --list

SKILL.md

Frontmatter

{
    "name": "smoke-test",
    "description": "Quick e2e smoke test of the local Carbon ERP dev server — logs in via \/auth, then loads each core module and verifies it renders without errors. Use after booting a stack, after wide-reaching changes, or when asked to \"smoke test\" the app. For feature-specific testing use \/test instead."
}

smoke-test — do the core modules load?

Verify every core module renders. This is breadth, not depth — it catches broken routes and crashed loaders, not logic bugs.

Announce at start: "Using the smoke-test skill — sweeping the core modules."

Prerequisites

Dev server running (crbn up — seeds the test user and DEV_BYPASS_EMAIL)

Steps

1. Login

Invoke /auth. If it fails, STOP and report — everything below needs auth.

2. Visit each module

Read ERP_URL from .env.local. For each route below:

agent-browser open ${ERP_URL}<path> && agent-browser wait --load networkidle && agent-browser snapshot -i

Module	Path
Dashboard	`/x`
Sales	`/x/sales/orders`
Purchasing	`/x/purchasing/orders`
Inventory	`/x/inventory`
Items	`/x/items/parts`
Accounting	`/x/accounting/charts`
People	`/x/people/employee`
Resources	`/x/resources`
Production	`/x/production`
Settings	`/x/settings/company`

A module passes if the snapshot shows real content (table, cards, headings) and no error text. It fails on a blank page, an error message, or a failed load. On failure: invoke /error to capture diagnostics, then continue to the next module — never abort the sweep for one failure.