Claude Code 2.1.152: Code Review Fixes, Skill Reloading, Vim Search

TL;DR

• /code-review --fix now applies findings directly to your working tree
• Skills can disable specific tools via frontmatter; new /reload-skills command re-scans without restart
• Vim mode: / in NORMAL mode triggers reverse history search; multiple bug fixes for thinking display, plugins, and MCP servers

New

• /code-review --fix — applies review findings to your working tree after analysis, surfacing reuse, simplification, and efficiency suggestions
• Skill tool disabling — skills and slash commands can set disallowed-tools in frontmatter to remove tools while the skill is active
• /reload-skills command — re-scans skill directories without restarting the session
• SessionStart hook improvements — hooks can return reloadSkills: true to re-scan directories and set session title via hookSpecificOutput.sessionTitle
• MessageDisplay hook — lets hooks transform or hide assistant message text as it displays
• pluginSuggestionMarketplaces setting — admins can allowlist org marketplaces for context-aware plugin suggestions
• Marketplace scope flags — claude plugin marketplace remove now accepts --scope user|project|local
• Fallback model switching — Claude Code switches to --fallback-model for the rest of the session when primary model is unavailable
• Auto mode consent removed — no longer requires opt-in
• Vim reverse history search — / in NORMAL mode opens reverse history search (like Ctrl+R in bash/zsh)
• /usage large file scanning — breakdown now includes large session files with streaming reads to keep memory flat
• Thinking summary display — collapsed group summaries stay readable for 3+ seconds, render as markdown, cap at 10 lines (Ctrl+O shows full)
• Fullscreen thinking indicator — counts up live while model thinks, preserves value if interrupted
• Workflow tool progress — simplified inline display; live agent counts show only in persistent status row
• Background agent timer — shows "Waiting for N background agents/workflows" and reports cumulative time when finished
• OpenTelemetry entrypoint metric — added app.entrypoint attribute (opt-in via OTEL_METRICS_INCLUDE_ENTRYPOINT=true)

Fixed

• Terminal styling degradation in very long sessions — renderer style pool now recycles properly
• Sandbox-enabled warning not appearing in condensed startup mode — now shows in every layout
• Loading spinner showing "still thinking"/"almost done thinking" while tool runs — thinking status now resets to "thinking" after each tool
• Focus mode spurious "N messages hidden" count on turns with no hidden activity
• Clicking links inside expanded tool results collapsing the section instead of opening the link
• Markdown table cell borders inheriting inline code color; wrapped continuation lines losing style; empty header cells showing labels in narrow-terminal stacked layout
• Plugin MCP servers with same command but different environment variables being incorrectly deduplicated
• /doctor reporting "marketplace not found" or "plugin not found" for stale enabledPlugins entries
• Plugins tracking git branches silently stopping updates after registry rebuild
• Remote MCP servers failing to connect in Claude Code Remote sessions with egress proxy enabled
• Effort-change confirmation dialog appearing with no messages or when switching between equivalent effort levels
• Agent tool description referencing undelivered agent list with --bare or attachments disabled
• Background worker crash in claude agents when accepting stale permission prompt after subagent cancellation
• cache_creation_input_tokens reporting as 0 in transcript when API reports cache writes via nested breakdown
• PushNotification tool incorrectly reporting "Mobile push not sent" in SDK-hosted sessions with Remote Control enabled
• Sessions stuck after model/login switch left stale thinking-block signatures — now stripped proactively with retry safety-net

Update Claude Code: npm install -g @anthropic-ai/claude-code@latest

Source: Claude Code