Guides
Can you share Claude skills across a 3D team?
Share Claude skills across a 3D or game team: commit them to the repo, run a private plugin marketplace, upload to claude.ai, and keep them safe.
6 min read

Quick answer
Can I share Claude skills with my 3D team?
Yes. The simplest way is to commit the skill folders to your project repository under .claude/skills, so everyone using Claude Code gets them on the next pull. For a wider set, package them as a Claude Code plugin marketplace. claude.ai users upload the same folders as custom skills, since skills don't sync between surfaces.
Why share skills at all
A skill on one artist's laptop helps one artist. The same skill in the project repository means every Claude on the team exports with the same FBX settings, names assets the same way and runs the same QA checks. That's the whole value of a pipeline, and it's where skills stop being a personal productivity trick and start being studio infrastructure.
There are three ways to share, from simplest to most structured.
Option 1: commit them to the project
Claude Code reads project skills from .claude/skills/<skill-name>/SKILL.md inside the repository (Claude Code skills docs). If your Blender assets or game live in git, this is all it takes:
- Put each skill in
.claude/skills/in the project repo. - Commit and push.
- Teammates pull. Their Claude Code sees the skills the next time it starts in that folder.
What you get for free:
- Review. Changes to export settings arrive as pull requests, with a diff anyone can read.
- History. When an export breaks,
git logon the skill shows what changed. - Scope. Skills stay with the project they're written for. The mobile game's budgets don't leak into the PC project.
For big repos, nested skills help: a skill in characters/.claude/skills/ loads the first time Claude works on a file in that folder. The character team's rigging checks stay out of the environment artists' way.
Option 2: a private plugin marketplace
When several projects need the same skills, such as a studio-wide naming convention or the QA checklist, copy-pasting folders between repos gets messy. Claude Code's plugins solve this. A plugin keeps its skills at <plugin>/skills/<skill-name>/SKILL.md, and a marketplace is a repository that lists plugins.
- Create a repository for studio skills, structured as a plugin marketplace. Anthropic's public skills repository shows the layout.
- Each artist adds it once:
/plugin marketplace add <your-org>/<repo>. - They install the plugins they need, for example
/plugin install asset-standards@<marketplace>.
Now one fix to the shared QA skill reaches every project that installed it, instead of five separate copies.
Option 3: claude.ai for people who don't use Claude Code
Not everyone on a 3D team works in a terminal. Concept artists and producers may use claude.ai or Claude Desktop. There, custom skills are uploaded rather than read from disk, and they need a Pro, Max, Ultra, Team or Enterprise plan with code execution on.
Skills don't sync between surfaces. Treat the git copy as the source of truth and re-upload when it changes. Put a version line in each skill ("Version 2026-11-21") so anyone can see whether their uploaded copy is current. Differences between the surfaces are covered in Claude skills in claude.ai vs Claude Code.

A repo layout that scales
For a studio with a few projects, this layout keeps things findable:
studio-skills/(the marketplace repo), with one plugin per area:asset-standards,asset-qa,export-unity,export-unreal,generation-intake- each game repo's
.claude/skills/, only for rules that apply to that game alone, such as its budgets or a special naming exception CLAUDE.mdin each game repo, with two lines pointing at the plugins the project expects
When a project-level skill turns out to be useful elsewhere, promote it to the marketplace and delete the local copy, so there's never two versions of the same rule.
What to share, and what to keep personal
Share
- engine export recipes
- naming and scale standards (building a standards skill)
- the asset QA checklist (building a QA skill)
- scene templates and render presets
- intake rules for generated models: budgets, clean-up steps and which generator settings to use
Keep personal
Put these in ~/.claude/skills/:
- shortcuts that match one person's habits
- experiments you haven't tested on real files
- anything that references one machine's paths
Rules that keep shared skills safe
Skills can contain scripts, and a shared skill runs on everyone's machine. Anthropic's guidance is to treat installing a skill like installing software. For a team, that becomes a few house rules:
- Review every change. Nobody merges a skill change, script or description, without a second pair of eyes.
- No secrets in skills. API keys go in environment variables or the tool's own sign-in, never in
SKILL.mdor a script. - No surprise network calls. If a script downloads or uploads something, the skill says so in plain words.
- Read-only by default. QA and reporting skills shouldn't modify files. Anything that writes or exports gets
disable-model-invocation: true, so it runs only when someone types the command. - Audit third-party skills before adding them to the studio marketplace, as carefully as you'd audit a Blender add-on.
Keeping skills from rotting
Skills go stale when Blender, the engine or the rules change. A light process works:
- Give each shared skill an owner.
- Re-test after each Blender or engine upgrade. The
bpyAPI changes between releases, and a script that worked in one version can fail in the next. - Keep a short changelog at the bottom of
SKILL.md. - When someone corrects Claude in chat more than twice for the same thing, that correction becomes a skill change.
Sharing tool access, too
Skills are half of a shared setup. The other half is the tools they call:
- Blender MCP runs locally, so each artist installs the add-on and registers the server. Document the steps in the repo's README. The guide to copy from is Blender MCP with Claude.
- Remote MCP servers, such as the Vi3W MCP server for generating, rigging, retexturing and cleaning up models, are added once per person as a custom connector in Claude. On Team and Enterprise plans, an owner adds the connector and members click Connect.
A shared skill can then say "generate the prop in Vi3W, run Clean shape to the budget, then import into Blender", and it works the same for everyone.
What sharing doesn't solve
Shared skills don't make everyone's Claude identical. Model choice, permissions and what each person has connected still differ. Shared skills also don't replace a human lead who decides what the standards should be. They make sure the standards, once decided, get applied.
Start with one
Pick the skill your team would argue least about, usually export settings. Commit it, and ask everyone to use it for a week. To give the team a steady supply of test assets, set up Vi3W and connect its MCP server to Claude.
Frequently asked questions
claude skills github
Anthropic publishes example skills at github.com/anthropics/skills. You can add that repo in Claude Code with /plugin marketplace add anthropics/skills. A private GitHub repo works the same way for your team's own skills.
can i sell claude skills
A skill is a folder of files you wrote, so you can license it like other text and code you own. Anthropic's docs don't describe a paid store for skills. Buyers should still read every file before installing, because skills can run scripts.
claude skills marketplace
In Claude Code, a marketplace is a repository that lists plugins, which can contain skills. Add one with /plugin marketplace add <owner/repo> and install plugins from it. A studio can run its own private marketplace for internal skills.
Last reviewed

