Claude Code v2.1.149: Usage Breakdown, Permission Fixes, and Keyboard Navigation

Anthropic released Claude Code v2.1.149, a maintenance release with several new features, a new enterprise setting, and numerous bug fixes — notably patching a PowerShell permission bypass that could allow reading outside the workspace.

New Features

• /usage now shows a per-category breakdown of what drives your limits: skills, subagents, plugins, and per-MCP-server cost.
• /diff detail view can be scrolled with the keyboard: arrows, j/k, PgUp/PgDn, Space, Home/End.
• Markdown output now renders GFM task list checkboxes (- [ ] and - [x]) instead of plain bullets.
• Enterprise: Added the allowAllClaudeAiMcps managed setting to load claude.ai cloud MCP connectors alongside managed-mcp.json.

Security Fixes

• PowerShell permission bypass: Built-in cd functions (cd.., cd\, cd~, X:) changed the working directory undetected, letting a later command read outside the workspace. Fixed.
• Sandbox write allowlist in git worktrees: Previously covered the entire main repository root; now scoped to only the shared .git directory (with hooks/ and config denied).
• PowerShell prefix/wildcard allow rules (e.g. PowerShell(dotnet.exe build *)) were not pre-approving native executables and scripts. Fixed.
• Permission-analysis gap: Parser trusted stale variable-tracking values for PWD/OLDPWD/DIRSTACK across cd/pushd/popd. Fixed.
• find in the Bash tool exhausted the macOS system file/vnode table and crashed the host on large directory trees. Fixed.

Other Notable Fixes

• Managed-settings approval dialog no longer leaves the terminal frozen after accepting at startup.
• /ultraplan and remote session creation no longer fail with “Could not capture uncommitted changes” when the working tree has no real changes.
• otelHeadersHelper failures (e.g., script path with spaces) are now reported in /doctor and the debug log.
• Thinking spinner no longer stays amber across tool calls and onto fresh thinking bursts.
• Collapsed Bash output now reports the correct hidden-line count for outputs with many short lines.
• Slash-command argument-hint no longer clips trailing typed characters when the hint overflows the input box.
• Argument-hint and progressive arg suggestions now appear after Tab-completing a skill whose frontmatter name: differs from its directory basename.
• Status bar now shows the effort level applied by skill/agent effort: frontmatter, not the baseline /effort setting.
• Ctrl+O transcript view now tails new messages instead of freezing at the moment it was opened.
• Editing a recalled prompt-history entry no longer loses the edit when navigating with arrow keys.
• /config exit summary no longer reports phantom changes to auto-compact and theme when toggling unrelated settings.
• /insights no longer crashes when cached session-meta files are missing optional fields.
• Malformed PowerShell and History tool calls with missing input are no longer misclassified as reads in transcript collapsing.
• Renaming a Remote Control session from claude.ai or the mobile app now updates the local session name for claude --resume.
• Fixed a race where a just-submitted prompt could appear twice in the up-arrow history.
• Tapping the “Jump to bottom” pill in fullscreen mode now dismisses it immediately.
• /feedback reports now include the conversation that happened before context compaction, making issues from earlier in long sessions easier to triage.