Labsco
Labsco/Repos/vercel-labs/openreview
REPO PACKAGE

vercel-labs openreview

COMMUNITY
vercel-labs · publisher1,449 repository starsFreegithub.com/vercel-labs/openreview
7skills
5ready to use

The seven skills are next-best-practices (file conventions, RSC boundaries, data patterns), next-cache-components (Next.js 16's Cache Components, cacheLife, cacheTag), next-upgrade (version upgrades via official codemods), vercel-composition-patterns and vercel-react-best-practices (React composition and performance guidance from Vercel Engineering), vercel-react-native-skills (React Native and Expo performance), and web-design-guidelines (a UI review against Vercel's own Web Interface Guidelines). None of the seven mentions OpenReview, pull requests, or code review directly — they're the same reference material the bot's Claude-powered reviewer consults when it comments on a diff.

This serves two different readers. One wants an automated GitHub reviewer: deploy the app to Vercel, register a GitHub App with contents/issues/PR read-write access, and mention @openreview in a PR comment to trigger a review, fix, and push. The other just wants these seven skills as a standalone Next.js and React reference and never needs to run the bot at all — six of the seven need nothing beyond the skill file itself, and one needs a local tool already installed.

READ THE FULL ANALYSIS

The bot itself is explicitly beta. The README calls it an internal project built to help the Vercel team test their own technologies together and says plainly to expect rough edges and breaking changes — a caveat about the review workflow, not about the seven skills sitting inside it.

Running the bot costs more than reading the skills. It needs a Vercel deployment, an Anthropic API key, a registered GitHub App with a private key and webhook secret, and optionally a Redis URL for persistent state — without Redis it falls back to in-memory, which does not survive a restart. None of that is required to copy a skill file into your own project instead.

5Work with nothing else to set up.
1,449Stars on the GitHub repository, at last check.