I Built a Plugin Because Anthropic Won't Stop Shipping

Key Takeaways

• Claude Code updates released daily.
• Frequent changes can break custom hooks.
• /whats-new plugin filters release notes to config.
• Shows only relevant changes since last check.
• Install with two simple Claude plugin commands.

Summary

Anthropic’s Claude Code is updated every day, delivering fixes and new features but also introducing breaking changes that can cripple custom hook configurations. Developer Brad Feld built a /whats-new plugin that scans a user’s Claude Code setup—hooks, rules, skills, commands, plugins, and environment variables—and cross‑references it with the latest release notes. The plugin categorises changes into directly relevant, potentially useful, and irrelevant, showing only the items that affect the user’s configuration since the last review. Installation requires just two Claude plugin commands.

Pulse Analysis

AI coding assistants are entering a phase of hyper‑iteration, with companies like Anthropic pushing daily updates to Claude Code. While rapid releases accelerate feature delivery, they also create a volatile environment for developers who have built intricate automation layers around the model. Missing a subtle change to hook execution or settings schema can halt pipelines, force emergency fixes, and erode confidence in the tool. This churn underscores a growing need for smarter change‑management solutions that surface only the most pertinent information to end‑users.

Enter the /whats-new plugin, a lightweight extension that bridges Claude Code’s public release notes with a developer’s personal configuration. By parsing hooks, rules, skills, commands, plugins, and environment variables, the plugin maps each change to the exact components a user relies on. It then sorts updates into three tiers: immediate impact, potential enhancements, and generic noise. Users can query the plugin without arguments to see changes since their last review, or specify a version to dive deeper. The result is a concise, actionable digest that eliminates the need to wade through exhaustive changelogs, allowing engineers to focus on building rather than firefighting.

Beyond the immediate convenience, the plugin illustrates a broader trend in the AI tooling ecosystem: lead users are creating niche utilities that address gaps left by rapid product cycles. Drawing on Eric von Hippel’s lead‑user theory, such contributions often become de‑facto standards, prompting platform owners to integrate community‑driven solutions into official marketplaces. As AI assistants become core components of software development stacks, tools that automate release‑note triage will be essential for maintaining velocity and ensuring reliable deployments across enterprises.