Added /teleport and /remote-env slash commands for claude.ai subscribers to resume and configure remote sessions. Added support for disabling specific agents using Task(AgentName) syntax in settings.json permissions or --disallowedTools CLI flag. Added hooks support to agent, skill, and slash command frontmatter with PreToolUse, PostToolUse, and Stop hooks. Added new Vim motions including ; and , for repeating f/F/t/T motions, y operator for yank, p/P for paste, text objects, >> and << for indent/dedent, and J to join lines. Added /plan command shortcut and slash command autocomplete support when / appears anywhere in input. Added --tools flag for interactive mode to restrict built-in tools. Added CLAUDE_CODE_FILE_READ_MAX_TOKEN_TOKENS environment variable and once: true config for hooks. Added YAML-style lists in frontmatter allowed-tools field. Added support for prompt and agent hook types from plugins. Added Cmd+V for image paste in iTerm2, left/right arrow navigation in dialogs, real-time thinking block display, and Skills as separate context visualization category.
Fixed OAuth token refresh, session persistence, session resume failures, race condition in token refresh, AWS Bedrock subagents EU/APAC configuration, API context overflow, FIFO file hang, terminal keyboard mode reset, Alt+B/F word navigation, ${CLAUDE_PLUGIN_ROOT} substitution, file permissions, bash command parsing, slash command execution, Tab-completion, command hint display, redundant Skill tool invocation, skill token estimates, subagent model inheritance, model picker, duplicate Bash commands, background task output, notifications, forked slash commands, cursor visibility, hooks menu, queued prompt images, pasted image size, CJK multiline prompts, ultrathink highlighting, reading indicators, Bash read command counts, spinner token counter, git diff memory leak, LSP tool race condition, feedback submission hang, search mode behavior, hook success messages, and improved startup performance.
Improved terminal rendering performance, Jupyter notebook reading, piped input reliability, AskQuestion tool, sed in-place edits, automatic continuation on token limit, compaction reliability, subagent permission handling, skills progress display, skill visibility in slash command menu, skill suggestions prioritization, spinner feedback, token count display, incremental async output, permission prompt UX, Claude in Chrome notification, macOS screenshot paste, and /stats output. Updated Atlassian MCP integration, changed "Interrupted" message color, removed permission prompt for plan mode, removed underline from image links. SDK and VSCode platform improvements.
Fetched April 11, 2026