/simplify Looks Like It Came From Someone's Dotfiles

Anthropic shipped two new Claude Code skills: /simplify (code review) and /batch (automated code migration using agents working in parallel). /batch is fine. /simplify is the interesting one, and not entirely in a good way.

GitHub Copilot does it. Cursor does it. CodeRabbit, Sourcegraph Cody, Qodo (all of them have some version of this). When Anthropic ships their own take, the implicit pitch is that Claude’s judgment is better than the others. Maybe. But that’s not the part worth thinking about.

What makes it weirder: Anthropic already ships a /code-review plugin for Claude Code. Four parallel agents, confidence-based scoring, full PR review. So /simplify isn’t just entering a crowded third-party market. It’s overlapping with something Anthropic themselves already built.

/simplify doesn’t solve a key problem on its own: it has no idea what your team’s coding standards are unless they’re in your CLAUDE.md files. Out of the box, it’s doing generic review. Catch the obvious stuff, flag patterns it was trained to dislike. Which is fine. But if your standards are in CLAUDE.md… Claude was already reading those files on every session. So what exactly is /simplify catching that Claude wasn’t already in a position to catch?

BUDDY: They packaged the thing Claude was already doing into a skill, named it, and shipped it. That’s not a feature. That’s a label.

I don’t think the answer is “nothing.” There’s probably real value in the structured context that /simplify sets up: forcing Claude to look at a diff with review as the explicit job, rather than as background context. Task framing changes what models pay attention to. That part holds up.

But it feels less like a strategic bet on code review as a competitive surface and more like Anthropic sharing how their own engineers use Claude Code internally. The way /remote-control and the Cowork announcements felt like Anthropic staking out the personal agent space. This doesn’t feel like that. Which is genuinely useful. I’ll use /simplify. I just wouldn’t read it as a move. It reads more like a post-it note from an internal wiki that got cleaned up and shipped.

I’m still adding both to my CLAUDE.md setup this week. But when the bar is “does this feel like Claude Code making a claim on something,” this one doesn’t clear it.