Anthropic: Claude Code v2.1.152 Brings /code-review --fix, /simplify and disallowed-tools in Skills Frontmatter

Anthropic released Claude Code v2.1.152, an update to the AI-assisted programming CLI that brings a range of practical improvements for everyday development workflows. Three key changes affect the productivity of teams using Claude Code in production development cycles.

How does /code-review —fix apply findings to the working tree?

The /code-review command received a new --fix flag that changes its behavior: instead of only reporting code review findings, Claude Code now automatically applies the discovered improvements directly to the project’s working tree. The focus is on reuse, simplification, and efficiency suggestions — the same category of findings that /simplify targets.

The /simplify command now internally calls /code-review --fix, unifying the two commands into a consistent workflow. Developers who used /simplify for code cleanup automatically get the same result with the same mechanism as the --fix flag.

What does disallowed-tools in skills frontmatter change for skill development?

Skills and slash commands in Claude Code now support a new disallowed-tools field in the YAML frontmatter. By setting this list, a skill author can remove specific tools from the model while the skill is active — enabling more precise context management and preventing unwanted actions for specific tasks.

Example: a skill intended exclusively for code review can use disallowed-tools to prohibit file-writing tools or shell command execution, ensuring Claude acts only as an analytical tool in that context.

Alongside disallowed-tools, /reload-skills was introduced — a new command that scans skill directories without restarting the session. SessionStart hooks can now return reloadSkills: true for automatic rescanning at session start, and can set the session title through hookSpecificOutput.sessionTitle.

New hooks and other changes

The new MessageDisplay hook event allows hooks to transform or hide assistant text at display time — useful for filtering sensitive content or reformatting responses in specific work environments.

Fallback model support: Claude Code now automatically switches to the configured --fallback-model for the remainder of the session if the primary model is unavailable, without interrupting work. Auto mode no longer requires opt-in consent from the user.

Vim mode gained / in NORMAL mode for searching command history (equivalent to Ctrl+R). Thinking summaries remain visible for at least 3 seconds, are rendered as Markdown, and are limited to 10 lines — Ctrl+O shows the full content. Fixes include terminal styling degradation in long sessions, incorrect reporting of cache_creation_input_tokens as 0, and stabilization of claude agents background workers.