Claude Code Digest — 2026-03-26 08:22:57
What the docs reveal
Anthropic is dismantling the monolithic structure of Claude Code. Today’s documentation updates introduce a formal claude plugin CLI command and deprecate the deeply entrenched /review function. The code review capability remains, but you must now install it explicitly via claude plugin install code-review@claude-plugins-official. This change shifts Claude Code from a batteries-included utility to a modular extension ecosystem. Maintaining specialized tools inside the core binary bloats the software and slows down iteration. By decoupling auxiliary functions from the main runtime, Anthropic can update specific AI workflows without forcing a global CLI upgrade.
You must now assemble your own toolchain. Breaking out /review into an independent package disrupts established workflow muscle memory. Developers who rely on quick, automated code audits will hit an error the next time they run the command. You must pause and install the official plugin first. However, this immediate friction buys long-term adaptability. The explicit @claude-plugins-official tag establishes a formal registry architecture. This structure proves Anthropic is clearing the runway for third-party developers, and eventually, internal corporate registries. Engineering teams will soon build and distribute proprietary context-gathering tools without waiting on Anthropic for native support.
The documentation updates also enforce strict namespacing for all custom skills. A developer building a new command can no longer claim a generic route; you must use the format /my-first-plugin:hello. This restriction signals Anthropic’s expectation for a sprawling user-generated ecosystem. Without namespaces, plugin collisions run rampant. If two downloaded extensions attempt to define a generic /test or /build command, the agent risks executing the wrong script. Typing fully qualified namespaces requires more keystrokes, but explicit execution guarantees predictable behavior.