What problem does it solve? When working with Vercel through the official Vercel MCP server, agents often pick the wrong tool, pull excessive data, or answer factual questions from memory instead of documentation. This Skill teaches the agent a disciplined procedure for using the connected Vercel MCP: which tool to call, at what depth, and in what order. ## Core Features & Use Cases - Depth ladder: Four escalating analysis levels from quick_status to debug_deployment_issues, so only the minimum sufficient data is fetched (token economy). - Tools-to-prompts map: A complete catalog of Vercel MCP tools (list_deployments, get_deployment_build_logs, get_runtime_logs, deploy_to_vercel) and workflow prompts (fix_recent_build, project_health_check, troubleshoot_common_issues). - MCP-first, CLI-fallback routing: Vercel CLI is used only when no MCP tool exists or a tool call fails. - Anti-hallucination docs routing: Factual Vercel questions are routed to search_vercel_documentation instead of answered from memory. - Use Case: A build fails on Vercel. The agent starts with quick_status, escalates to get_deployment_build_logs only if needed, localizes the failure, and proposes a fix — requesting human confirmation before any redeploy. ## Quick Start Ask the agent why your latest Vercel build failed and have it diagnose the issue using the Vercel MCP tools.
