Settings menu supports left and right arrow keys to switch options and save instantly
When you want to adjust tool preferences without repeatedly pressing confirm across dropdown menus, there is now a faster way. Typing /config in the terminal opens the settings page. Highlight the setting you want to change and press ← or → on your keyboard to instantly switch to the previous or next value and save immediately, without needing to press Enter to expand the dropdown list. The footer at the very bottom of the screen also displays a ←/→ Change prompt.
FEATURE
Fullscreen chats no longer freeze, with smoother mouse selection and repainting
When you resize terminal windows or select text with a mouse during long chats, the display responds much more quickly. In the flicker-free altscreen (terminal fullscreen display mode) and artifact viewer, dragging the window will no longer freeze the interface. The system prioritizes repainting the messages currently on screen instead of getting stuck at the old width waiting for the entire conversation to finish repainting. Pressing Esc to interrupt the agent or resume the conversation also avoids repainting all steps on screen from scratch. Dragging to highlight past the top or bottom with the mouse now auto-scrolls, and scrolling the mouse wheel or selecting text while the agent streams output will not cause selections to drift. Copying text also captures lines scrolled off screen and automatically strips line numbers and annotation previews in the artifact viewer.
FEATURE
Simpler Go command approvals allow permanent authorization for builds and checks
When you frequently have the agent compile or check Go project code, the authorization process is much easier. When approving safe commands like go vet or go build, the system now provides an option to allow all future arguments for that subcommand, so passing different arguments later will not require manual confirmation again. However, for subcommands that execute or install code—such as go run, go test, go generate, go install, and go tool—approvals must still cover the exact full command.
FEATURE
Queued messages support merged immediate delivery, fixing Vim shortcut mishandling
When you enter multiple commands while the agent is busy, message queuing and recall are now more reliable. When enabling Send Immediately in /config, if messages are already queued, newly submitted input is delivered to the agent in order along with the existing queued messages rather than falling behind in a separate queue. If message delivery fails mid-turn, the message automatically returns to the queue instead of vanishing. Pressing k on the top line in Vim normal mode now pulls queued messages back into the editor like the Up arrow key, fixing an issue where pulling history mistakenly caused duplicate submissions of queued messages.
FEATURE
Image generation moves to a dedicated subagent with automatic retries and artifact saving
When you ask the agent to draw illustrations or design assets, the image generation workflow is now better organized. The primary agent delegates image creation to the built-in image-generator subagent. It handles prompt writing, inspects each generated image, and makes up to three attempts automatically until successful before saving the completed images to the conversation artifacts. The entire generation process is clearly recorded in the conversation transcript as an independent subagent execution stage.
BUGFIX
Interrupted background commands no longer hang, with accurate waiting in headless mode
When long-running background commands executed by the agent are terminated by the system or run out of memory, execution no longer stalls. Previously, if a background command crashed or was killed by the system due to out-of-memory errors, the agent remained stuck waiting. The system now correctly recognizes exit codes like 137 or 134 and proceeds with subsequent tasks. When running in headless mode with the -p flag, it also fixes an issue where the process exited prematurely if a background command finished after the first turn before the agent responded.
When you launch a session from deep within a project subdirectory, project-level customizations now take full effect. Previously, starting a session from a subdirectory ignored custom definitions like skills.json and rules.json in parent .agents/ folders. The system now traverses upward from the current working directory to the project root, loading configuration files across all levels. The menu opened by typing / also fixes an issue where it listed built-in skills disabled for the current agent, preventing selection from injecting instructions for tools the agent cannot use.
BUGFIX
Configuration files with UTF-8 BOM now supported without Windows saving errors
When editing configuration files on Windows using Notepad or PowerShell's Set-Content, editors often prepend a UTF-8 byte order mark (a hidden character indicating text encoding). Previously, this prevented ~/.gemini/config/config.json from loading settings properly. This issue has now been resolved, ensuring smooth parsing across various editors.
BUGFIX
Fixed word wrapping and indentation glitches when terminal lines are exactly full
When slash command output or system warnings match the exact width of the terminal window, the final word previously wrapped to the start of the next line while losing indentation and disrupting the layout. The rendering logic is now fixed, maintaining clean wrapping and consistent indentation even when text reaches the exact window boundary.
Original changelog
Original changelogSettings menu supports left and right arrow keys to switch options and save instantly
Added `←`/`→` shortcuts to `/config`: on a highlighted setting they switch to the previous or next value and save it right away, without opening the dropdown, and the footer shows a `←/→ Change` hint
Original changelogFullscreen chats no longer freeze, with smoother mouse selection and repainting
Improved responsiveness in long conversations in no flicker (altscreen) mode: resizing the terminal no longer freezes or lags the CLI, the full-screen view redraws the messages on screen first instead of showing text wrapped for the old width until the whole conversation has re-rendered, and interrupting the agent with `Esc` or resuming a conversation no longer re-renders every step already on screen.
Improved mouse text selection in the `altscreen` view and in the artifact viewer: dragging past the top or bottom now auto-scrolls, the selection stays on its text while the agent streams or you scroll with the wheel, and copying includes lines that scrolled off screen, without line numbers or comment previews from the artifact viewer
Original changelogSimpler Go command approvals allow permanent authorization for builds and checks
Improved the always-allow suggestion when approving `go` commands: approving a command such as `go vet` or `go build` now offers to allow that subcommand with any arguments, while `go run`, `go test`, `go generate`, `go install`, and `go tool` still require the exact command
Original changelogQueued messages support merged immediate delivery, fixing Vim shortcut mishandling
Improved the `Send Immediately` option for Queued Messages in `/config`: a message you send while earlier messages are still queued now goes to the agent together with them, in order, instead of joining the queue, and a message that fails to send mid-turn goes back to the queue instead of being lost
Fixed `k` in Vim Normal mode recalling the last history entry instead of your queued messages when pressed on the top line, which could send a queued message twice; `k` now pulls queued messages back into the editor like the `Up` arrow does
Original changelogImage generation moves to a dedicated subagent with automatic retries and artifact saving
Changed how the agent generates images: it now hands image requests to a built-in `image-generator` subagent, which writes the prompt, checks each result with up to three attempts, and saves the images to the conversation's artifacts, so image generation appears as a subagent run in the conversation
Original changelogInterrupted background commands no longer hang, with accurate waiting in headless mode
Fixed the agent stalling when a background command it was waiting on crashed or was killed by the system, for example when it ran out of memory; the agent now sees the command finish with its exit code, such as `137` or `134`, and keeps going
Fixed headless `-p` runs sometimes exiting before the agent could respond to a background command that finished after its first turn
Original changelogSubdirectories properly load parent skills, and slash menus omit disabled tools
Fixed `skills.json`, `rules.json`, and other customization manifests in a parent `.agents/` directory being ignored when a session started in a subdirectory; manifests now load from every `.agents/` directory between the working directory and the project root
Fixed the `/` menu listing built-in skills that the current agent does not enable, which inserted instructions for tools the agent could not use when selected
Original changelogConfiguration files with UTF-8 BOM now supported without Windows saving errors
Fixed settings failing to load when `~/.gemini/config/config.json` starts with a UTF-8 byte order mark, as files saved by Notepad or PowerShell `Set-Content` on Windows do
Original changelogFixed word wrapping and indentation glitches when terminal lines are exactly full
Fixed slash-command output and alerts where a line exactly as wide as the terminal had its last word pushed to the start of the next line without indentation