Claude Code v2.1.152 Major Updates

Claude Code v2.1.152 Major Updates

Claude Code v2.1.152 was released on May 26, 2026. This update introduces new features including automatic working tree application for `/code-review --fix` and skill management, along with 16 bug fixes.
2026.05.27

This page has been translated by machine translation. View original

This is Ishikawa from the Cloud Business Division. Claude Code v2.1.152 was released on May 26, 2026. This update includes new features related to automatic application of code review results to the working tree and skill management, along with numerous bug fixes.

Update Summary

v2.1.152 includes 33 changes: 8 new features such as /code-review --fix and skill management (/reload-skills, disallowed-tools), multiple usability and stability improvements, and 16 bug fixes at the core. There are no breaking changes or deprecations.

Note that v2.1.151 was not released (skipped), and the preceding v2.1.150 contained only internal improvements with no user-facing changes. Therefore, v2.1.152 is the most recent release with substantive changes.

Highlighted Updates

  • New Feature: /code-review --fix now directly applies suggestions for reuse, simplification, and efficiency to the working tree after a review. Additionally, the /reload-skills command and disallowed-tools specification for skills/slash commands have been added, improving skill management operability.
  • Bug Fix: An issue where sessions would become unresponsive after switching models or logging in has been resolved.

Update Details

New Features

  • /code-review --fix working tree application: /code-review --fix now directly applies suggestions for reuse, simplification, and efficiency to the working tree after a review. Additionally, /simplify now internally calls /code-review --fix (as an alias).

https://dev.classmethod.jp/articles/20260528-code-review-fix/

  • disallowed-tools front matter: You can now specify disallowed-tools in the front matter of skills and slash commands, allowing you to exclude specific tools from the model while that skill is active.
  • /reload-skills command: A /reload-skills command has been added that allows you to rescan the skills directory without restarting the session.
  • Skill reloading via SessionStart hook: When the SessionStart hook returns reloadSkills: true, skills installed by the hook become available within the same session.
  • Session title setting via SessionStart hook: At startup or resumption, the SessionStart hook can now set the session title via hookSpecificOutput.sessionTitle.
  • MessageDisplay hook event: A new hook event MessageDisplay has been added, allowing you to transform or hide the text of displayed assistant messages.
  • pluginSuggestionMarketplaces admin setting: An admin setting that allows administrators to allowlist organization marketplaces for plugins suggested in context-aware hints.
  • OpenTelemetry attribute app.entrypoint: The session entrypoint can now be added as an OpenTelemetry metrics attribute (opt-in via OTEL_METRICS_INCLUDE_ENTRYPOINT=true).

Improvements

  • Automatic switching to --fallback-model: Instead of failing every request when the primary model is not found, the session now switches to the configured --fallback-model for the remainder of the session.
  • Auto mode opt-in no longer required: Auto mode can now be used without prior opt-in consent.
  • /usage support for large session files: Large session files are now included in the /usage breakdown. Files are read via streaming, so memory usage remains constant.
  • Improved thinking summary display: Thinking summaries within collapsed groups now remain readable for at least 3 seconds, are rendered as markdown, and are truncated at 10 lines (press Ctrl+O to view the full content).
  • Live display of "Thinking for Ns": In fullscreen mode, the thinking time display now counts up live while thinking is in progress and retains the value even if interrupted midway.
  • Simplified Workflow tool progress display: The live agent count is now displayed only in the persistent workflow status line below the prompt.
  • Waiting display for background processing: The post-response timer now indicates that it is waiting when background agents/workflows are running, and reports the cumulative time after results have been processed.
  • --scope support for marketplace remove: claude plugin marketplace remove now accepts --scope user|project|local, bringing it in line with add/install/uninstall.
  • Vim mode history search: In NORMAL mode, / now opens a reverse history search (same as Ctrl+R), matching bash/zsh vi-mode behavior.

Fixes

The main bug fixes are as follows:

  • Fixed session becoming unresponsive: Fixed an issue where old thinking block signatures remained in history after switching models or logins, causing sessions to hang (signatures are now proactively removed, and a retry safety net has been added).
  • Fixed terminal display degradation in long sessions: Resolved an issue where terminal styles degraded during long sessions by reusing the renderer's style pool.
  • Fixed sandbox-enabled warning not displaying: Fixed an issue where the sandbox-enabled warning was not shown in the compact launch display; it now appears in all layouts.
  • Fixed incorrect deduplication of plugin MCP servers: Fixed an issue where plugin MCP servers with the same command but different environment variables were incorrectly deduplicated.
  • Fixed git branch tracking plugin stopping updates: Fixed an issue where a plugin tracking git branches would stop receiving updates after the plugin registry was rebuilt.
  • Fixed remote MCP connection failure: Fixed an issue where remote MCP servers would fail to connect in Claude Code Remote sessions when an egress proxy was enabled.
  • Fixed background crash in claude agents: Fixed an issue where a background worker would crash when an old permission prompt was accepted after a sub-agent was cancelled.
  • Fixed cache_creation_input_tokens showing as 0: Fixed an issue where cache_creation_input_tokens appeared as 0 in transcript and result usage when the API reported cache writes only within nested cache_creation breakdowns.
  • In addition, numerous minor bugs have been fixed, including loading spinner display, markdown table rendering, false reports in /doctor, the effort change confirmation dialog, and PushNotification notification determination.

Closing

v2.1.152 is an update that adds new features such as working tree application for /code-review --fix and skill management capabilities, along with fixes for session hangs and issues related to MCP and plugins. Since there are no breaking changes or deprecations, you can update without worrying about impact on existing workflows. If you regularly use skills, plugins, or code review features, this update has a broad impact on your workflow, so why not update and give it a try?

Recent Updates

https://dev.classmethod.jp/articles/20260524-claude-code-updates-v2-1-150/

https://dev.classmethod.jp/articles/20260528-code-review-fix/

References

https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md

https://code.claude.com/docs/en/changelog


Claudeならクラスメソッドにお任せください

クラスメソッドは、Anthropic社とリセラー契約を締結しています。各種製品ガイドから、業種別の活用法、フェーズごとのお悩み解決などサービス支援ページにまとめております。まずはご覧いただき、お気軽にご相談ください。

サービス詳細を見る

Share this article