Wondering what the Claude Code skill doctor actually does — and whether it’s worth running? The quick version is just below. Anthropic added the new /skill-doctor command in Claude Code v2.1.261, released 4 September 2026 per the official Claude Code changelog, and it answers a question I hear constantly: which of my skills are quietly eating my context window without ever doing anything?
Short answer:
- What it is:
/skill-doctorshows which loaded skills go unused and what they cost in context, so you can prune them — shipped in Claude Code v2.1.261 on 4 September 2026, per the official changelog. - Where the report lives: in an interactive session it opens in the
/pluginmanager’s Stats tab; in headless mode with-pit prints as plain text, per the official docs. - What it flags: skills that have never been invoked (and where to turn them off), plus plugins you haven’t used recently.
- Requirements: Claude Code v2.1.252 or later, run from the terminal where the session lives — it’s not available over Remote Control connections.
- Why bother: every stale skill costs context tokens on every request. Pruning them leaves more room for the work you actually want done.
What Is the Claude Code Skill Doctor?
The Claude Code skill doctor is a built-in audit command for your skills. Skills are brilliant — they’re how I’ve packaged up my SEO workflows, content checklists and publishing routines inside Claude Code — but they have a hidden price: every skill you keep loaded takes up space in the context window on your requests, whether you use it or not.
Until now, working out which skills were pulling their weight meant guesswork. The changelog entry for v2.1.261 (4 September 2026) puts it plainly: /skill-doctor was added “to show which loaded skills go unused and what they cost in context, so you can prune them”.
The official Claude Code docs go further: the report shows what each of your skills costs and how often it gets used, flags skills in the listing that have never been invoked, tells you where to turn them off, and also lists plugins you haven’t used recently. One caveat straight from the docs — the report covers the skills in your session other than bundled skills and enterprise skills, so those won’t appear in the listing.
If you’ve been stacking up skills the way most power users have this year — a deploy skill here, a review checklist there, half a plugin marketplace on top — this is the spring clean the tool has needed.
How to Use the Claude Code Skill Doctor
Using the Claude Code skill doctor takes about ten seconds:
- Update Claude Code. The docs say the feature requires v2.1.252 or later, and the command shipped in v2.1.261 — so get current first.
- Run
/skill-doctorinside your session. In an interactive session, the report opens in the/pluginmanager’s Stats tab. - Headless? In non-interactive mode with
-p, Claude Code prints the report as text — handy if you want to pipe it into your own tooling or log it as part of a scheduled job. - Act on it. For anything flagged as never invoked, the report says where to turn it off. The
/skillsmenu lets you toggle skill visibility withSpaceand save the change to your settings, and/reload-pluginsapplies plugin changes without a restart.
Two limits worth knowing before you go hunting for the command, both from the official docs: it isn’t available in sessions that skip feature-flag fetching, and it isn’t available over Remote Control connections from your phone or browser — you have to run it in the terminal where the session is running.
🔥 Want this set up without the guesswork? Inside the AI Profit Boardroom we run Claude Code skill audits like this as part of a full AI SEO system — 3,700+ members, four live calls per week, daily tutorials, done-for-you templates and a 30-day roadmap, so you’re never left wondering which skills or workflows are worth keeping.
Prefer to talk it through one-on-one first? Book a free SEO strategy session and we’ll map out where Claude Code fits in your content pipeline.
What the Report Actually Tells You
Three things, and each one maps to a decision:
| Report signal | What it means | What to do |
|---|---|---|
| Context cost per skill | How much of your window that skill’s listing consumes | Question anything expensive that you rarely invoke |
| Usage frequency | How often the skill actually gets invoked | Keep the workhorses; audit the passengers |
| Never-invoked flag | The skill has literally never fired in your sessions | Turn it off — the report tells you where |
| Stale plugins list | Plugins you haven’t used recently | Disable or uninstall via /plugin |
This matters more than it sounds. Context is the one budget every Claude Code session spends whether you like it or not. If you’re running content workflows — briefs, drafts, internal linking, publishing — the space wasted on a skill you installed in March and never touched again is space your actual instructions and page content can’t use. I’ve covered the same principle from another angle in my guide to Claude Code prompt cache metrics: measure what your setup costs before you scale it.
Why This Fits an SEO Workflow
Most people reading this aren’t pruning skills for fun — they’re running Claude Code as a content and SEO machine. Here’s the practical chain: leaner skill listings mean more room for page drafts, keyword data and site context in each request; more room means fewer truncated or compacted sessions mid-task; and fewer interruptions mean batches of content actually ship.
The rest of the v2.1.261 release backs the same theme of controlling what fills your context. The changelog notes new bashOutputMaxChars and taskOutputMaxChars settings that raise how much command and background-task output Claude receives inline before it’s saved to a file — up to 128K characters — plus a new --append-subagent-system-prompt-file flag for reading a subagent system prompt from a file when it’s too large for the command line. Different levers, same idea: you decide what deserves the window.
If you’re building out a serious setup, pair the skill audit with a look at Claude Code managed MCP servers — MCP tool definitions occupy context too — and, if you run locked-down environments, my walkthrough of Claude Code restricted mode.
The Bottom Line on the Claude Code Skill Doctor
The Claude Code skill doctor is one of those small features that quietly pays for itself every single session. It shipped on 4 September 2026 in v2.1.261, it takes one command to run, and it turns “I think I have too many skills” into a concrete list: what each one costs, what never fires, and where to switch it off. Run it once, prune hard, and re-run it whenever you bolt on a new plugin. Your context window is the most expensive real estate in your whole AI stack — stop renting it out to skills that don’t work for a living.
And if you want the full system — not just the pruning but the skills worth keeping — that’s exactly what we build together inside the AI Profit Boardroom.
FAQ: Claude Code Skill Doctor
What does the Claude Code skill doctor do?
It reports what each of your loaded skills costs in context and how often it gets used, flags skills that have never been invoked, and lists plugins you haven’t used recently — so you can decide which ones to turn off. It was added in Claude Code v2.1.261 on 4 September 2026, per the official changelog.
How do I run /skill-doctor?
Type /skill-doctor inside a Claude Code session. Interactively, the report opens in the /plugin manager’s Stats tab; in headless mode with -p, it prints as plain text.
Which version of Claude Code do I need?
The docs state v2.1.252 or later is required, and the command itself appears in the v2.1.261 changelog entry. If the command isn’t recognised, update Claude Code first.
Does the report cover every skill?
No — per the official documentation it covers the skills in your session other than bundled skills and enterprise skills.
Why does /skill-doctor say it is unavailable in my session?
Two documented limits: it’s not available in sessions that skip feature-flag fetching, and it’s not available over Remote Control connections from a phone or browser. Run it in the terminal where the session actually lives.
Does removing unused skills actually improve Claude Code’s output?
Skill listings consume context on your requests, so pruning skills you never invoke frees that space for the task itself — which is precisely the problem this command was built to expose.
Related reading
Next step: if you want your Claude Code setup audited, pruned and pointed at content that actually ranks, join 3,700+ members inside the AI Profit Boardroom — four live calls a week, daily tutorials, templates and a 30-day roadmap. Or book a free SEO strategy session and I’ll help you map the quickest path for your site.
About the author: Julian Goldie is an SEO agency owner with 394K+ YouTube subscribers, a 100% job-success score on Upwork, 75K+ community members across his groups, 10+ years in SEO and a best-selling SEO book. Catch his latest tutorials on YouTube, learn AI SEO with him inside the AI Profit Boardroom, or book a free SEO strategy session. For agency work, book a call for a custom quote.
Last updated September 2026. This is the living guide to claude code skill doctor — it gets updated as the tools change.
