Anthropic / github.com/anthropics/claude-code Claude Code v2.1.152: /code-review --fix Applies Changes Directly, Skills Get Disallowed-Tools
Anthropic shipped Claude Code v2.1.152 on May 27, adding --fix to the code review command so suggestions land in your working tree automatically. The release also gives skills and slash commands a way to restrict the tool list while they're active.
Anthropic published Claude Code v2.1.152 in the early hours of May 27. The headliner is /code-review --fix, which goes from flagging issues to actually applying the fixes. There’s also a new tool-restriction system for skills, a live reload command, and a hook for transforming what Claude displays.
/code-review —fix
Previously, /code-review ran an analysis and showed you findings. You still had to apply them yourself. With --fix, Claude applies the suggested changes to your working tree after completing the review. The release notes say it surfaces reuse, simplification, and efficiency suggestions in addition to correctness issues.
As a side effect, /simplify now just calls /code-review --fix. The two commands do the same thing now.
Skills Can Restrict the Tool List
Skills and slash commands can now include disallowed-tools in their frontmatter. While that skill is active, the listed tools are removed from the model’s available set. That means you can write a focused review skill that prevents Claude from reaching for Bash, or a documentation skill that locks out file-editing tools entirely.
Combined with the skill-specific tool lists that already existed, you now have both sides of the access control: tools to add, and tools to remove.
/reload-skills
/reload-skills rescans your skill directories without restarting the session. If you’re editing skills while Claude Code is running, you no longer have to kill and reopen the process to pick up changes.
SessionStart hooks can also return reloadSkills: true to trigger the same rescan on startup and resume.
Session Title From Hooks
SessionStart hooks can now set the session title via hookSpecificOutput.sessionTitle. The title appears in the agents view and session history. Useful if you’re running Claude Code programmatically and want descriptive labels in the agents panel instead of the default timestamp-based names.
MessageDisplay Hook
The new MessageDisplay hook fires when Claude is about to display assistant text. You can use it to transform or suppress the output before it renders. The primary use case in the release notes is enterprise deployments where certain content should be filtered or reformatted.
Auto Mode Drops the Opt-In
Auto mode no longer requires an opt-in consent step. It’s active without any prompt once you’re on a plan that supports it. The auto mode classifier still handles permission decisions in the background.
Vim: / for Reverse History Search
In Vim NORMAL mode, pressing / now opens reverse history search (equivalent to Ctrl+R in non-Vim mode). A small addition, but one that had been tripping up people who use Vim everywhere and kept accidentally triggering the wrong behavior in Claude Code’s terminal.
Sources: Release v2.1.152 — Anthropic on GitHub, May 27, 2026; Claude Code changelog — Anthropic