Claude Code stopped auto-running /verify and /code-review — if you relied on them firing implicitly, your quality gate silently disappeared

Claude Code Docs·medium signal

Version 2.1.215 (July 19) changed Claude Code so it no longer invokes the `/verify` and `/code-review` skills on its own; you now call them explicitly. This is a small line in the changelog with a large practical consequence: any workflow that assumed a review pass happened automatically at the end of a task now ships unreviewed by default. The fix is to make the invocation explicit where it matters — a hook, a slash command in your task template, or a line in CLAUDE.md/AGENTS.md that tells the agent to run `/code-review` before declaring done.