Claude Code 2.1.228 patches a conversation-history leak in Remote Control

Claude Code 2.1.228 fixes a Remote Control bug: running /resume while a remote device was connected could leak the resumed conversation’s title or history into the connected session. Several other fixes go to the self-hosted runner and cross-session messaging.
The release notes on GitHub, published on August 11, list 18 entries, and 11 of them are fixes. One entry changes how a tool behaves: the Write tool now lets newer models overwrite an existing file they haven’t read in the current session, under the same rules the Edit tool uses. Older models still have to read the file before they write it.
According to the Claude Code docs, Remote Control connects claude.ai/code or the Claude app for iOS and Android to a session running on your own machine. The web and mobile apps are a window onto that local session, and the conversation stays in sync across every connected device.
The self-hosted runner, which arrived in version 2.1.224 as a way to run web, mobile and desktop sessions on your own machines or containers, already has two fixes. Sessions failed on every fresh runner when the checkout hook failed for a repository the session never pushes to. That repository is now skipped with a warning. Runners also ended sessions in the gap between a background task finishing and the follow-up turn starting.
Cross-session messaging gets a fix and a display change. The first session after an install or upgrade sometimes started without an inbox. Messages now show the sender and body inline instead of on one collapsed line, and a message sent to a Remote Control session on another machine shows your Remote Control session name as the sender.
Two fixes deal with deleted data. Session cleanup was removing the contents of a project’s memory folder. Background plugin-cache cleanup was deleting a plugin’s cache when its only version was a symlinked development checkout.
The rest are smaller:
- interactive sessions could stop redrawing after a rare internal layout error while the process kept running
- on Windows, git and Git Bash weren’t found when Claude Code was launched from a parent folder of the git installation
/tuireverted the session to an earlier model if/modelhad changed since the last response- a marketplace entry redefined in a higher-precedence settings tier could inherit another tier’s custom headers; entries now merge as whole entries
- the deferred-tools reminder was sometimes sent to the model twice after a skill invocation
- expired or missing Google Cloud credentials on Vertex AI now fail within seconds instead of retrying for minutes
- compaction now shows the retry countdown and the stall hint, where before there was only a progress bar
- new busy-spinner glyphs in the terminal title reduce tab-bar jitter on some terminals
- the first-use notice for Pro, Max and Team plans no longer says auto mode sessions cost slightly more
Skills synced from claude.ai were also hardened. They can no longer shadow local commands or MCP prompts, and their descriptions are now sanitized and labeled. When one of these skills runs on your machine, its body won’t execute ! commands or expand @ file references.
Up next
Grok Bot beta gives every agent its own computer in the cloud, Video
xAI opened an early beta of Grok Bot: users spin up their own agents for real tasks, and each bot gets its own cloud computer plus login access to the services and apps you already use. It’s live for SuperGrok Heavy, Cursor Ultra and Cursor Teams Premium.
Related stories
- Claude Code 2.1.229: plugins load from a directory with no restart
- Claude Code 2.1.227 stops pitching credits to Max subscribers
- Claude Code sessions can now run on your own machines
- Claude Code on desktop folds an iPhone Duo to test layouts
- Claude Code 2.1.318: login expiry warning, runner log change
- Claude Code 2.1.312 stops overwriting big CLAUDE.md files
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

