Claude Code v2.1.261: /skill-doctor, Larger Inline Output Limits, and Bash-Style Word Editing

Claude Code v2.1.261 is a large release with 67 changelog entries, 26 of them for VS Code. It adds /skill-doctor for finding unused skills, settings that raise how much command output Claude receives inline, a diagnostic line that explains why an organization policy failed to load, and Bash-style word editing in the prompt. The remaining fixes focus on Remote Control, session resume, input handling, and provider setup on Bedrock and Vertex AI.

What’s New

`/skill-doctor`

/skill-doctor shows which loaded skills go unused and what they cost in context, so you can prune them.

Larger Inline Output Limits

The new bashOutputMaxChars and taskOutputMaxChars settings raise how much command and background-task output Claude receives inline before it is saved to a file, up to 128K characters.

Organization Policy Diagnostics

/status and claude doctor now include an “Organization policy” line that says why your organization’s policy could not be loaded, such as a proxy not passing the endpoint through.

Subagent System Prompt from a File

--append-subagent-system-prompt-file reads the subagent system prompt from a file, for prompts too large to pass on the command line.

Bash-Style Word Editing

The prompt’s word-editing keys now match Bash: Ctrl+W deletes back to whitespace, Alt+F and Alt+D stop at word end, and punctuation separates words. keybindingFlavor no longer has any effect.

Auto Mode and Safety Prompts

Auto mode now treats a link that packs content into a public diagram renderer’s URL as an upload to that site, so it is no longer auto-approved unless you asked for it.
The dangerous-rm safety prompt now also catches rm -rf on positional parameters and inside double-quoted sh -c scripts.

Machines whose managed settings pin forceLoginMethod: "gateway" now ignore a leftover API key or claude.ai login and ask for /login; Bedrock, Vertex AI, and Foundry sessions are unaffected.
A Claude apps gateway 403 on the managed settings load (at startup or after /login) now says Claude Code may not be enabled for the organization, instead of advising a new sign-in.

Model Names in Pickers

When Claude Code recognizes a model, the /model picker and the VS Code model pill now show its name instead of its raw Bedrock, Vertex AI, or LLM gateway ID. In VS Code, the model picker is now one flat list of every model, with rows kept for older model spellings listed last.

VS Code

The Output styles menu adds a “Build a custom style” walkthrough that writes a custom output style file and lists it right away.
The MCP servers dialog adds an Add server form and a Remove action, so MCP servers can be added and removed without leaving the IDE.
Sessions open in a terminal, another VS Code window, or Claude Desktop now show a hollow ring in the session list, so they no longer look closed.
Permission and question prompts get a fold button, so the conversation behind them can be read without dismissing them; the space beside the prompt now scrolls the conversation.
“Archive session” is now in the session list’s right-click menu, and Unarchive has its own icon.

Fixes and Improvements

Input and terminal

Typed or pasted characters no longer occasionally land out of order or get dropped during fast input or key repeat.
You can now delete the character immediately before an inline [Image #N] chip in the prompt input.
The terminal progress indicator (iTerm2, Ghostty, ConEmu) no longer shows the session as finished while a background workflow or agent is still running.

Sessions and resume

Resuming a session no longer loses hook output and other context around parallel tool calls, which had changed the resumed request.
SDK and cloud sessions no longer ignore a Stop or interrupt sent just after the first prompt, before the turn has started; the turn now stops instead of running to completion.
claude -p --resume <file> no longer adopts a malformed session ID recorded in the transcript; it now resumes under a fresh session ID instead.
Sustained high CPU usage when a background agent could not be resumed and its wake-up was retried in a tight loop is fixed.

Remote Control

Remote Control no longer shows a stale permission mode when a phone, browser, or claude.ai app attaches to a terminal session, or after the mode changes in the terminal.
Remote Control sessions no longer show as still working (stuck spinner and Stop button) after stopping a turn from a connected phone or browser, or after a local slash command like /clear.
SendMessage to an offline Remote Control session on another machine no longer reads as delivered; the result now says delivery is queued until that machine reconnects.
Remote Control’s inbound event stream no longer fails behind TLS-inspecting corporate proxies on native Windows.

Providers and connectors

The Bedrock setup wizard no longer hangs when AWS or an AWS credential helper never responds (it now times out with a clear error), and its model checks no longer fail behind a TLS-inspecting proxy.
On Google Vertex AI with GOOGLE_APPLICATION_CREDENTIALS set, startup is faster: API client creation no longer re-runs Google Cloud project discovery or spawns extra gcloud processes.
gcpAuthRefresh no longer opens a browser at startup when the Google credential check is slow but the credential is still valid.
claude.ai connectors no longer stay absent for the whole session when the startup connector fetch times out; the CLI now retries in the background.

Prompt cache and performance

In-process agent-team teammates no longer re-send their first-turn tool and skill announcements on the second turn, which had changed the request prefix and missed the prompt cache.
Streaming performance is improved: already-rendered blocks are no longer re-checked by layout on each update.
When the API sends no response headers, the retry now waits up to API_TIMEOUT_MS (10 minutes by default) instead of another 3 minutes, and the messages say what to change.
/context token counting now uses a local estimate when the token-counting API is unavailable, instead of extra small-model requests.

v2.1.261 contains 67 changes in total; see the full changelog for the complete list.

How to Update

Check your installed version: claude --version
Update in place: claude update
If you set keybindingFlavor, note that it no longer has any effect; the prompt’s word-editing keys now match Bash.