Vercel Platform

**Quick judgement** A strong, comprehensive Vercel Platform skill that exposes most of the Vercel CLI surface (deployments, projects, domains, env vars, logs, teams) plus direct access to Vercel documentation in markdown. It is powerful enough for serious CI/CD and operations workflows, but also carries meaningful production-risk if granted broad write access without guardrails. **What it’s good for** - Standing up and iterating on applications hosted on Vercel (from `vercel dev` to `vercel --prod`). - Managing environment variables across dev/preview/prod and syncing them into local `.env` files. - Operational tasks on existing deployments: inspecting, viewing logs (including JSON for `jq`), promoting to production, rolling back, and redeploying. - Domain and alias management, including buying or transferring domains, setting aliases, and managing SSL/DNS. - On-demand access to Vercel docs and sitemap as markdown, which is ideal for an AI agent needing authoritative, up-to-date platform details. **Key risks and limitations** - **Destructive operations**: Commands like `vercel rm`, `vercel projects remove`, `vercel domains remove`, and misused `rollback`/`redeploy` can cause downtime or data loss if invoked incorrectly. - **Secrets and privacy**: Env var management and logs can expose secrets or sensitive runtime data if not handled carefully. - **Cost and configuration risk**: Domain purchases (`vercel domains buy`) and domain transfers can incur costs or misconfigure production DNS if not tightly controlled. - **Scope/team confusion**: Using `--scope`, `vercel switch`, or `vercel teams` incorrectly could apply operations to the wrong team or project. **Recommended usage scenarios** - As a controlled deployment and ops tool for teams already standardized on Vercel, especially where an AI agent is assisting with routine deploys, log inspections, and rollbacks. - In CI/CD pipelines or dev tooling where commands are templated and reviewed, limiting the chance of destructive, ad-hoc actions. - For documentation retrieval and platform exploration (using the docs endpoints and sitemap) in a low-risk, read-only mode when you don’t want the agent modifying real infrastructure. Given the breadth of write-access operations, this skill is best used with **clear policies and guardrails**, e.g., allowing an agent to read logs, inspect deployments, and fetch docs freely, but requiring human approval for destructive or high-impact commands (remove, domain changes, production deploys).