Claude Code v2.1.269 introduces new features for plugin evaluation and several updates aimed at improving the development environment.

New Features and Improvements

  • Addition of claude plugin eval: You can now run plugin evaluation suites and obtain results in JSON and HTML report formats.
  • Agent Map Visualization in VSCode: Allows for listing sub-agents within a session, viewing them in individual cards, stopping agents, and reviewing read-only transcripts.
  • Switchable Output Styles: Through /output-style [name], users can now select and change output styles for remote control, cloud, and headless sessions.
  • Improved Progress Display in VSCode: A live progress line for running sub-agents now appears under tool call groups.
  • Optimized Prompt Suggestions for Multilingual Support: Suggestion filtering has been improved for languages without spaces between words, such as Japanese, Chinese, and Thai.

Bug Fixes

  • Prompt Cache Issue: Fixed an issue where prompt cache was partially invalidated when a response was interrupted by output token limits and then automatically resumed.
  • Session Resumption Behavior: Fixed an issue where the method of sending context during session resumption changed, reducing cache reusability.
  • VSCode UI Issues: Fixed several bugs related to session list display, plan preview latency, and tab behavior.
  • Other Fixes: Resolved issues including CMYK JPEG decoding errors, key input malfunctions in certain terminal environments, and behavior during Slack integration.

Sources (official release notes):