TL;DR
Shift+Tab in permission comment fields was granting session-wide edit permission; now fixed (more below). If you've used Shift+Tab to navigate out of a comment during a permission prompt, it was silently approving the edit and granting blanket edit access for the session.
New opt-in
spellchecksetting highlights misspelled words as you type in the prompt input. Requiresaspell,hunspell, orispellinstalled; enable it in settings to get real-time spell checking.Permission dialog "don't ask again" now correctly reflects the actual grant scope. The displayed scope matches what accepting would cover, and the option is withheld when contents cannot be fully shown.
New in 2.1.235
2.1.235 (August 19, 2026)
- Added an optional
spellchecksetting that underlines misspelled words in the prompt input as you type, using your installedaspell,hunspell, orispell - Fixed whole-prompt-cache invalidation when a language server disconnected or reconnected mid-session
- Fixed nested markdown list items misaligning at depth 3+ and added a hanging indent to wrapped list items in the terminal UI
- Fixed prompt input highlights (slash commands, keywords, mentions) appearing shifted by one or more characters in some multi-line prompts
- Fixed Shift+Tab inside the permission prompt's comment field approving the edit and granting session-wide edit permission instead of closing the field
- Fixed the Agent tool advertising a general-purpose default in sessions where that agent is unavailable: an omitted
subagent_typethere now gets a clear error listing the available agents - Fixed notebook cell delete/replace approval dialogs silently omitting the existing cell content when the notebook or cell could not be read; the dialog now says why
- Fixed slash commands run while Claude is responding showing HTML entities instead of the actual characters
- Fixed the prompt footer not showing the "Update installed" restart notice after a background auto-update
- Fixed the expanded task list (
ctrl+t) always starting collapsed when resuming or relaunching into a session that still has open tasks - Improved memory and CPU usage while cloud sessions such as
/ultrareviewor/autofix-prrun in the background; their event streams are no longer re-scanned and re-rendered on every update - Improved permission dialogs: display text and "don't ask again" options now always match what a grant would cover, and "don't ask again" is withheld when contents cannot be fully displayed
- Improved the embedded
grepin native macOS/Linux builds: pathological patterns now fail fast instead of exhausting memory, and-m Nwith-A/-Cprints correct context - Improved the context-limit error to say when auto-compact is off and point to
/configto re-enable it - Vim mode: NORMAL mode and cursor position are now preserved when toggling the detailed transcript (ctrl+o) or closing a panel
- Dialogs: arrow keys and Enter pressed in quick succession now select the option you navigated to instead of the previously highlighted one
SendMessagenow refuses messages too large for cross-session delivery up front instead of silently dropping them- Remote Control:
claude rcnow applies the same enterprise-gateway availability check as interactive startup - [VSCode] Fixed focus jumping between open Claude tabs on its own when a window with several Claude panels is restored or reloaded
Notes
Permission dialogs: a steady hardening arc
2.1.235 closes a particularly sharp gap: Shift+Tab in a permission comment field was indistinguishable from accepting the prompt. Combined with the "don't ask again" scope fix in the same release, both the accidental-approval and the over-scoped-approval vectors are tightened in one build.
This continues a pattern across recent releases. 2.1.234 fixed permission answers being dropped when approving background subagent prompts. 2.1.232 added Bash input-redirection permission checks and fixed a PowerShell variable-writing bypass. 2.1.223 closed a bypass where crafted commands could hide parts from the approval dialog. If you relied on Shift+Tab to close a comment field without acting on the prompt, that expectation needs updating.