Personal vs Project Skills in Claude Code (2026)
A. Frans
Published October 19, 2026
Table of Contents
- 01All seven skill locations
- 02Personal skills: `~/.claude/skills/`
- 03Project skills: `.claude/skills/` in the repo
- 04Nested skills for monorepos
- 05Plugin skills
- 06Which copy wins when names collide
- 07Do you need to restart after adding a skill?
- 08A worked example: one developer, one monorepo
- 09A simple rule for where each skill goes
- 10FAQ
Claude Code reads skills from seven different places, and the folder you drop a SKILL.md into decides who gets it, which projects see it, and which copy wins when two skills share a name. Most people only know two of the seven: ~/.claude/skills/ for personal skills and .claude/skills/ inside a repo. That covers the common case, until a teammate's project skill silently loses to your personal one, or a skill in packages/api/ never shows up in the / menu.
This guide covers every location, the precedence rules, and a simple decision for where each kind of skill belongs. Everything here comes from Anthropic's current Claude Code skills documentation.
All seven skill locations
| Location | Path | Who gets it |
|---|---|---|
| Personal | ~/.claude/skills/<name>/SKILL.md | You, in every project on this machine |
| Project | .claude/skills/<name>/SKILL.md (repo root) | Anyone who opens this repo, once committed |
| Nested | <subdir>/.claude/skills/<name>/SKILL.md | Sessions started in or below that subdirectory, or once Claude touches files there |
| Additional directory | .claude/skills/ inside a folder passed with --add-dir | That session only |
| Plugin | <plugin>/skills/<name>/SKILL.md | Anyone with the plugin enabled, invoked as /plugin-name:skill-name |
| Enterprise | .claude/skills/ in the managed settings directory | Every user on machines where your organization deploys it |
| claude.ai account | Skills enabled on your claude.ai account | Cowork, cloud sessions, and terminal sessions signed in with that account |
~/.claude/skills/ do not load in Cowork or cloud sessions. They live on your disk, and those sessions don't. If you want a skill everywhere, enable it on your claude.ai account as well.
Personal skills: ~/.claude/skills/
This is your toolbox. Anything here loads in every repo you open on this machine. On Windows the same folder sits at C:\Users\<you>\.claude\skills\.
Put a skill here when it's about how you work, not about a particular codebase:
- Your commit message style
- A writing voice or an anti-filler editing pass
- A personal research or note-taking workflow
- Generic tools like the PDF skill or XLSX skill that you want available in any folder
mkdir -p ~/.claude/skills/commit-style
# then create ~/.claude/skills/commit-style/SKILL.md
Nobody else sees these. That's the point, and it's also the trap: if a project depends on one of your personal skills, the project breaks for everyone but you.
Project skills: .claude/skills/ in the repo
Project skills travel with the code. Commit .claude/skills/ and every teammate who clones the repo gets the same skills, at the same version as the code they describe.
Put a skill here when it's about this codebase:
- How to run this repo's test suite or deploy script
- The house style for this project's API endpoints
- A migration checklist for this database schema
- Release steps for this specific product
Claude Code also walks up the tree. It loads project skills from .claude/skills/ in the directory where you start it and in every parent directory up to the repository root. Start a session in packages/frontend/ and the root-level skills still load.
Committing skills also means they get code review. A skill can carry scripts and can pre-approve tools through the allowed-tools frontmatter field (explained in our SKILL.md frontmatter guide), so a pull request is a good place to catch a skill that quietly grants Bash(*).
Nested skills for monorepos
A skill in a subdirectory, say apps/web/.claude/skills/deploy/SKILL.md, behaves differently from a root skill. Per the docs, nested skills don't load at startup. They load the first time Claude reads or edits a file in that subdirectory, then stay available for the rest of the session. Until then they don't appear in the / menu, and you can't call them by name.
This is the single most common reason a monorepo skill "doesn't exist." If you need it right away, run /add-dir apps/web (Claude Code v2.1.257 or later) to load it immediately, or start the session inside apps/web/.
If a nested skill shares a name with a root skill, both load. You call the nested one by its qualified name, for example /apps/web:deploy.
Plugin skills
Skills bundled in a plugin live at <plugin>/skills/<name>/SKILL.md and get namespaced: a review skill in my-plugin becomes /my-plugin:review. The bare /review still works as long as nothing else claims that name.
Plugins are the right home when you want to distribute a set of skills beyond one repo, with versioning and one-command install. Anthropic's own document skills ship this way:
/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skills
Our explainer on Claude Code plugin marketplaces covers how to set one up for your team.
Which copy wins when names collide
| Collision | Result |
|---|---|
| Personal vs project, same name | Personal wins |
| Enterprise vs personal or project | Enterprise wins |
Skill vs a .claude/commands/ file of the same name | The skill wins |
| Root project skill vs nested skill | Both load; nested one uses its qualified name |
| Plugin skill vs anything above | Both load, because plugin skills are namespaced |
| Your skill vs a bundled command | Your skill replaces the command, not its aliases |
deploy in both ~/.claude/skills/ and the repo's .claude/skills/, typing /deploy runs your personal one, not the version your team committed. If you once wrote a personal deploy skill for a side project and then join a team that has its own, you'll keep running the wrong one without any warning.
Two habits prevent it:
1. Give personal skills a prefix that teams won't use, like my-commit or af-research. 2. When a project skill behaves oddly, check ~/.claude/skills/ for a same-named folder before debugging the skill itself.
If a skill doesn't trigger at all, the cause is usually its description, not its location. See why your Claude Code skill isn't triggering.
Do you need to restart after adding a skill?
Usually not. Claude Code watches ~/.claude/skills/, the project's .claude/skills/, and .claude/skills/ in any --add-dir directory, and picks up added, edited or removed skills within the current session.
Two exceptions:
- If you create a top-level skills directory that didn't exist when the session started (your first personal skill, for example), run
/reload-skills. - For a skill folder that's also a plugin, changes to
hooks/,.mcp.json,agents/oroutput-styles/need/reload-plugins.
One more trap: the permissions.additionalDirectories setting in settings.json grants file access only. It doesn't load skills from those directories. Use --add-dir or /add-dir for that.
A worked example: one developer, one monorepo
Here's how the locations combine for a developer working in a company monorepo with a web app and an API:
~/.claude/skills/
my-commit/SKILL.md # personal: commit message style
my-research/SKILL.md # personal: how I take notes
acme-monorepo/
.claude/skills/
run-tests/SKILL.md # project: how CI runs, for everyone
release/SKILL.md # project: release checklist
apps/web/.claude/skills/
deploy/SKILL.md # nested: Vercel deploy steps
services/api/.claude/skills/
migrate-db/SKILL.md # nested: schema migration steps
Start Claude Code at the repo root and you get my-commit, my-research, run-tests and release right away. deploy appears once Claude works on files under apps/web/, and migrate-db once it touches services/api/. Neither clutters the / menu for someone who only works on the other half of the repo. If someone later adds a root-level deploy, both load, and the web one is called by its qualified name, /apps/web:deploy.
The personal skills carry a my- prefix, so a teammate adding a project commit skill next month won't get shadowed on this machine.
A simple rule for where each skill goes
Ask one question: would this skill still make sense in a different repo?
- Yes, and only for me → personal (
~/.claude/skills/) - No, it describes this codebase → project (
.claude/skills/), committed - No, and only one package in a monorepo → nested (
<package>/.claude/skills/) - Yes, and the whole team or the public should get it → plugin
- Yes, and the whole company must get it → enterprise managed settings
If you're unsure, start personal. Promote the skill to the project or a plugin once a second person asks for it. Our guide to sharing agent skills with your team covers that move, and how to update, disable or remove skills covers cleanup afterward.
FAQ
Where are personal Claude Code skills stored?
In~/.claude/skills/<skill-name>/SKILL.md (on Windows, C:\Users\<you>\.claude\skills\). They load in every project on that machine but not in Cowork or cloud sessions.
Do project skills override personal skills?
No, it's the other way around. When a personal and a project skill share a name, the personal one runs. Enterprise skills override both.Why doesn't my monorepo skill show up in the slash menu?
Skills in a subdirectory's.claude/skills/ load only after Claude reads or edits a file in that subdirectory. Run /add-dir <path> to load them sooner, or start the session in that directory.
Should I commit .claude/skills/ to git?
Yes, if the skills describe the project. Committing them gives every teammate the same skills and lets you review them like any other code, including any tool permissions they grant.
Do I need to restart Claude Code after adding a skill?
Usually not. Claude Code watches skill directories and picks up changes mid-session. Run/reload-skills only if you created a top-level skills directory that didn't exist when the session started.Share this article
⚙Related Tools
📄Related Articles
Get More AI Tool Guides
New comparisons and guides every week. Join thousands of professionals staying ahead of the AI curve.