Agent Skills › ryparker/excisely › check-deployment

check-deployment

GitHub

验证 Vercel 部署健康状态，检查构建日志、运行时错误及页面可访问性，用于诊断部署失败或确认生产环境正常运行。

.claude/skills/check-deployment/SKILL.md ryparker/excisely

Trigger Scenarios

需要检查应用是否成功部署诊断部署构建失败原因验证生产环境功能是否正常

Install

npx skills add ryparker/excisely --skill check-deployment -g -y

More Options

Non-standard path

npx skills add https://github.com/ryparker/excisely/tree/main/.claude/skills/check-deployment -g -y

Use without installing

npx skills use ryparker/excisely@check-deployment

指定 Agent (Claude Code)

npx skills add ryparker/excisely --skill check-deployment -a claude-code -g -y

安装 repo 全部 skill

npx skills add ryparker/excisely --all -g -y

预览 repo 内 skill

npx skills add ryparker/excisely --list

SKILL.md

Frontmatter

{
    "name": "check-deployment",
    "description": "Verify that the Vercel deployment is healthy. Checks build status, runtime logs, and page accessibility. Use when you need to check if the app is deployed, diagnose deployment failures, or verify production is working."
}

Check Deployment

Verify the Vercel deployment is healthy using the Vercel MCP tools and CLI.

Instructions

1. Find the Project

Read .vercel/project.json to get the project ID and org ID:

cat .vercel/project.json

If that file doesn't exist, use the Vercel MCP:

mcp__claude_ai_Vercel__list_teams()  → get teamId
mcp__claude_ai_Vercel__list_projects(teamId)  → find the project

2. Check Latest Deployment

mcp__claude_ai_Vercel__list_deployments(projectId, teamId)

Look at the most recent deployment. Check its state:

READY = deployed successfully
ERROR = build failed
BUILDING = still in progress
QUEUED = waiting to build

3. If Build Failed — Get Logs

mcp__claude_ai_Vercel__get_deployment_build_logs(idOrUrl, teamId, limit: 200)

Read the logs to identify:

TypeScript errors
Missing environment variables
Dependency installation failures
Build timeout

4. If Build Succeeded — Verify Page Loads

mcp__claude_ai_Vercel__web_fetch_vercel_url(url: "https://<deployment-url>/")

Verify the page returns HTML, not an error page. Check for:

200 status
Expected page content (login page or dashboard)
No "Application error" or "500 Internal Server Error"

If you get a 403, generate a share link:

mcp__claude_ai_Vercel__get_access_to_vercel_url(url: "https://<deployment-url>/")

5. Check Runtime Logs for Errors

mcp__claude_ai_Vercel__get_runtime_logs(
  projectId,
  teamId,
  level: ["error", "fatal"],
  since: "1h",
  limit: 50
)

Look for:

Database connection errors
OpenAI API failures
Auth errors
Unhandled exceptions

6. Report

Summarize findings:

Deployment status (ready/error/building)
Build time
Any errors found in build or runtime logs
Whether the app is accessible and returning expected content

Quick Check (All-in-One)

For a fast health check, run these steps in order. Stop at the first failure and report the issue.

Tips

If environment variables are missing in production, they need to be set in the Vercel dashboard (Project Settings → Environment Variables)
Runtime logs are only available for serverless/edge functions, not static pages
Build logs are the most useful for diagnosing deployment failures