Claude Code v2.1.159 to v2.1.160 Major Updates

Claude Code v2.1.159 to v2.1.160 Major Updates

I will explain the main improvements, including security enhancements, background agent stabilization, and the change to the Dynamic Workflows trigger keyword (`workflow` → `ultracode`).
2026.06.02

This page has been translated by machine translation. View original

This is Ishikawa from the Cloud Business Division. Here is a summary of the Claude Code updates from v2.1.159 to v2.1.160 (May 31 – June 1, 2026).

Update Summary

This covers 2 versions (v2.1.159 – v2.1.160, May 31 – June 1, 2026). v2.1.159 includes only internal infrastructure improvements, with all user-facing changes consolidated in v2.1.160. The main highlights this time are security enhancements such as confirmation prompts before writing to shell startup files, stability improvements around background agents (claude agents), and a breaking change that renames the Dynamic Workflows trigger keyword from workflow to ultracode.

Notable Updates

  • Improvement: When Claude views a single file using grep/egrep/fgrep, a separate Read tool call before editing (Edit) is no longer required (relaxation of the read-before-edit check. v2.1.160).
  • Bug Fix: Fixed an issue where restoring completed sessions in claude agents or re-attaching background sessions after overnight retirement would cause chat history to be lost and the initial prompt to be re-executed (which could result in side-effectful operations being executed twice) (v2.1.160).
  • Security: Confirmation prompts are now displayed before writing to shell startup files, ~/.config/git/, and build tool configuration files (v2.1.160).

Target Versions and Period

Version Release Date Notes
v2.1.159 2026-05-31 Internal infrastructure improvements only (no user-facing changes)
v2.1.160 2026-06-01 Main body of user-facing changes

Security

Confirmation prompts have been added for writes that could lead to unintended command execution.

  • Confirmation before writing to shell startup files and Git configuration (v2.1.160): A confirmation prompt is now displayed before writing to shell startup files such as .zshenv, .zlogin, and .bash_login, or to files under ~/.config/git/. These files could potentially lead to unintended command execution if modified.
  • Confirmation before writing to build tool configuration even in acceptEdits mode (v2.1.160): Even in acceptEdits mode, which automatically approves edits, a confirmation prompt is now displayed before writing to build tool configuration files that could allow code execution (such as .npmrc, .yarnrc*, bunfig.toml, .bazelrc, .pre-commit-config.yaml, .devcontainer/, etc.).

Improvements

  • Files viewed with grep no longer require an additional Read before Edit (v2.1.160): Claude Code has a safety mechanism (read-before-edit check) that requires a file to be read before it can be edited (via Edit). This is a constraint on the agent's tool operations, not something the user manually executes with Read. Previously, even if the content was confirmed with grep, a separate Read tool call was required before editing. In v2.1.160, viewing a single file with grep/egrep/fgrep is now treated as satisfying this check, allowing the agent to proceed directly to editing right after confirming the relevant section with grep. As a result, unnecessary read steps are reduced and responses proceed more efficiently.
  • Background agent behavior improvements (v2.1.160): Performance when opening recently inactive sessions in claude agents has been improved. The latency of the auto mode classifier has also been improved, reducing how often routine operations are blocked with "could not evaluate this action." Additionally, during termination (claude rm/stop, idle collection), SIGTERM is now sent to running shell subprocesses before SIGKILL, ensuring that cleanup processes are properly executed.
  • Removed JetBrains plugin suggestion at startup (v2.1.160): The JetBrains plugin installation suggestion that was displayed at startup has been removed.

Fixes (Major)

Excerpting fixes related to stability and usability.

  • Fixed history loss when restoring completed sessions in claude agents (v2.1.160): Restoring a completed session would cause chat history to be lost and the initial prompt to be re-executed. This could lead to redoing already completed work or double-executing side-effectful operations.
  • Fixed conversation loss when re-attaching background sessions (v2.1.160): Fixed an issue where background sessions re-attached after overnight retirement would lose their conversation content and re-execute the initial prompt.
  • Fixed Japanese IME candidate display position (v2.1.160): Fixed an issue in the claude agents display where CJK IME composition text (such as Japanese) was appearing in the bottom-left of the screen instead of at the input position.
  • Fixed voice mode connection failure with non-ASCII paths (v2.1.160): Voice mode would fail to connect when the project directory name or branch name contained non-ASCII characters or special characters.
  • Fixed several-second freeze when returning to claude agents list (v2.1.160): Fixed an issue where the auto-updater would re-check every time the session list was returned to, causing a several-second freeze.
  • Fixed inability to input keys under high CPU load on Windows (v2.1.160): Fixed an issue on Windows where Esc, arrow keys, and input would become unresponsive under high CPU load while attached to a background session or viewing an agent display.
  • Fixed history loss when resuming in brief mode (v2.1.160): Fixed an issue where resuming a brief mode session with brief mode turned off would cause past replies to disappear from the scrollback buffer.
  • Fixed clipboard copy failure in WSL (v2.1.160): Fixed an issue where copy-on-select was not writing to the Windows clipboard in WSL. Switched from OSC 52 to PowerShell integration, enabling it to work with terminals that do not support OSC 52, such as MobaXterm.
  • In addition, numerous minor bugs have been fixed, including: startup failures in claude --bg due to "socket missing," deletion of session starting directories on Windows, incorrect "Completed" display in the agent list, display corruption in some terminals (Apple Terminal, tmux), broken file:///C:/... links, paste position in vim mode, and error messages with third-party providers.

Breaking Changes and Deprecations

  • Dynamic Workflows trigger keyword changed from workflow to ultracode (v2.1.160): The keyword for activating Dynamic Workflows has been changed from workflow to ultracode. As a result, the word "workflow" will no longer trigger a workflow (you can still use it by making your request in your own words). The trigger keyword is highlighted in violet in the prompt input field. As a related fix, /effort ultracode will no longer be displayed for models that cannot run xhigh, and the incorrect display attributing unavailability to Dynamic Workflows settings has also been resolved.
  • Removed environment variable CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE (v2.1.160): The environment variable has been removed and setting it will now be a no-op.

Closing

In this update, v2.1.159 consisted of internal improvements only, while v2.1.160 contained the user-facing changes. The main focus is security enhancements such as confirmation prompts before writing to shell startup files and build tool configuration files, along with stabilization of background agents. In particular, the change of the Dynamic Workflows trigger keyword from workflow to ultracode will affect those who have been using "workflow" to trigger it. If any of the changes catch your attention, go ahead and update to check them out.

Recent Updates

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

https://dev.classmethod.jp/articles/20260524-claude-code-updates-v2-1-152/

https://dev.classmethod.jp/articles/20260528-claude-code-updates-v2-1-153/

https://dev.classmethod.jp/articles/20260529-cc-updates-v2-1-158/

References

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

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


生成AI活用はクラスメソッドにお任せ

過去に支援してきた生成AIの支援実績100+を元にホワイトペーパーを作成しました。御社が抱えている課題のうち、どれが解決できて、どのようなサービスが受けられるのか?4つのフェーズに分けてまとめています。どうぞお気軽にご覧ください。

生成AI資料イメージ

無料でダウンロードする

Share this article