Engineering Den: Claude Code skills and hooks for frontend teams

We just shipped a shared skills and hooks repository for anyone on the UX or cloud team working on frontend features with Claude Code. One command installs everything into your repo, and from that point forward, your AI-assisted workflow has built-in guardrails with automatic updates to distribute.

Unlike typical pre-commit or pre-push hooks, these trigger inside Claude Code and Codex as you work without missing a beat. Finish an iteration, and they'll automatically check type safety, linting, formatting, testing, and builds. They also catch anti-patterns, like typing as any (effectively downgrading TypeScript to JavaScript), missing tests, browser-specific regressions, and accessibility gaps before they make it into a PR.

The real leverage is the lifecycle skill. It walks Claude Code through a structured development flow: ask clarifying questions, plan the work, write tests alongside the implementation, verify the result, open a PR, and generate a report you can share with the team. That report can land as a GitHub issue or a Jira ticket, depending on your setup. You can go very far with minimal hand-holding.

Lastly, experimenting with SandCastle: a community project that gives each Claude Code instance its own Docker container. Point it at a Jira ticket or GitHub issue and let it iterate on its own, with the same skills and hooks you'd have locally.

One command installs the skills for all collaborators in your repo. A verify script confirms everything is wired up correctly. It works with both Claude Code and Codex (and they can review each other's work).

Check it out at https://github.com/redpanda-data/ui-harness and drop feedback in the Redpanda Slack Community. 

‍