Agent skills explained: An FAQ
By Steven Van ·
Vercel's FAQ covers what agent skills are, how they differ from MCP servers and rules, and how to install a skill package.
Vercel published an FAQ explaining agent skills, a packaged, reusable way to give AI agents instructions, scripts, and resources they can load automatically for a given task.
The post frames skills as a fix for problems that show up once teams lean on agents for real workflows: prompt wording drifts between people, workflow conventions like quality checks and approval flows don't get inferred correctly, and pasting detailed playbooks into every prompt bloats the context an agent has to reason over. A skill package centralizes that context instead, built around a required SKILL.md file plus optional scripts, reference docs, and configuration files.
The FAQ also draws lines between skills and other pieces of agent setup:
- Skills package complete workflows, including when to use a tool, how to sequence actions, and what success looks like.
- MCP servers give agents standardized access to external tools and services.
- Tools are individual functions an agent can call for a single operation.
- Rules and system prompts set baseline behavior and constraints that apply across all tasks.
Examples cited include a Vercel React Best Practices skill for component performance patterns, a Supabase Postgres Best Practices skill for schema and query design, a copywriting skill for conversion-focused content, and a Remotion Best Practices skill for video generation. To install a package, files go in a project's skills/ directory or global user scope, or via a skills command-line utility.
Read the full agent skills FAQ on the Vercel blog.
