Anthropic · 2026-09-04 · major
Claude Code 2.1.261 — /skill-doctor shows which skills waste your context
Claude Code 2.1.261 adds /skill-doctor, which lists the loaded skills a session never used and what each one costs in context, so you can prune them. New settings raise inline command output to 128K characters.
Claude Code 2.1.261 adds a skill audit, much bigger inline output limits, and a stricter rm -rf safety check.
Key specs
| Max inline output | 128K characters |
|---|
Quick facts
| Maker | Anthropic |
|---|---|
| Version | 2.1.261 |
| Released | September 4, 2026 |
| New command | /skill-doctor (unused skills + their context cost) |
| New settings | bashOutputMaxChars, taskOutputMaxChars |
| New flag | --append-subagent-system-prompt-file |
| Install | npm i -g @anthropic-ai/claude-code@2.1.261 |
What is it?
A new /skill-doctor command in Claude Code 2.1.261 reports which loaded skills a session never used and what each one costs in context, so you can drop the ones that are only taking up room. The release, published September 4, 2026, also adds --append-subagent-system-prompt-file for subagent prompts too large to pass on the command line.
How does it work?
Two new settings, bashOutputMaxChars and taskOutputMaxChars, decide how much command and background-task output the model receives inline before it is written to a file; the ceiling is now 128K characters. Separately, /status and claude doctor gain an "Organization policy" line that says why a managed policy could not be loaded, such as a proxy that does not pass the endpoint through.
Why does it matter?
Two safety behaviours tighten in this build. The dangerous-rm prompt now also catches rm -rf on positional parameters and inside double-quoted sh -c scripts, and auto mode stops auto-approving a link that packs content into a public diagram renderer's URL, treating it as an upload to that site. Prompt word-editing keys also switch to Bash behaviour, which means the keybindingFlavor setting no longer has any effect.
Who is it for?
Claude Code users and the admins who manage them
Frequently asked questions
- How much command output can Claude Code 2.1.261 keep inline?
- Claude Code 2.1.261 lets you raise inline output to 128K characters using the new bashOutputMaxChars and taskOutputMaxChars settings. Below that limit the model receives command and background-task output directly in the conversation; above it, the output is saved to a file instead. The previous, lower ceiling meant long build logs were written out and had to be read back.
- Does Claude Code 2.1.261 change any keyboard shortcuts?
- Yes. Claude Code 2.1.261 changes the prompt's word-editing keys to match Bash: Ctrl+W deletes back to whitespace, Alt+F and Alt+D stop at the end of a word, and punctuation now separates words. Because that behaviour is fixed, the keybindingFlavor setting no longer has any effect and can be removed from your settings file.
- How do I find out why my organization's policy will not load?
- Run /status or claude doctor. Claude Code 2.1.261 adds an "Organization policy" line to both that explains the failure, for example a proxy that does not pass the policy endpoint through. The same release also changes a gateway 403 during the managed-settings load to say Claude Code may not be enabled for the organization, instead of asking you to sign in again.
- What changed for Remote Control in this release?
- Claude Code 2.1.261 fixes several Remote Control problems: stale permission modes when a phone or browser attaches to a terminal session, sessions that kept showing a spinner after a turn was stopped, a /teleport pull that was uploaded back into the connected session, and delivery status for offline machines, which now says a message is queued until that machine reconnects.
Try it
npm i -g @anthropic-ai/claude-code@2.1.261