One command installs skills across 75+ coding agents, and prints the prompt without installing
Synopsis
vercel-labs released skills, a CLI that installs agent skills into Claude Code, Codex, Cursor and 75 more coding agents with one command, and can emit a skill's prompt directly.
Interpretation
A skill package no longer has to be tied to one coding agent. A single command installs the same skill files into Claude Code, Codex, Cursor and more than 75 other agents, and public and private repositories use the same command. Previously a skill had to be distributed and installed separately for each agent, so switching agents meant redoing the work. skills add accepts GitHub shorthand, full GitHub URLs, direct paths to a skill inside a repo, GitLab URLs, Azure Repos URLs, any git URL and local paths, and private repositories reuse the authentication already configured for the repository URL.
A skill can also be used without installing it. skills use resolves sources the same way as skills add, writes the selected skill files to a temporary directory, and by default prints only the generated prompt to stdout. Previously, getting a skill's prompt meant installing it into an agent and extracting it by hand. With --agent, skills use starts one supported agent interactively with the generated prompt, as in npx skills use vercel-labs/agent-skills@web-design-guidelines | claude.
Installation can run fully non-interactively, which suits CI/CD. Previously, batch installation across several agents required confirming each one by hand. npx skills add vercel-labs/agent-skills --skill frontend-design -g -a claude-code -y sets the skill, global install, target agent and no-confirmation flag at once, and --all installs every skill in a repo to every agent.
Perspective
Intended for developers and engineering teams that need to reuse skill files across several coding agents, covering public and private repositories, GitHub, GitLab, Azure Repos, any git URL and local paths, with non-interactive installation for CI/CD. The material states no next step.
The material is a summary and gives no measured cross-agent compatibility results and no full list of the 75+ agents; when GitHub API access fails, update checks fall back to an authenticated git clone, and the material does not describe how that path performs.
