- In short
- There are four mechanisms for distributing a reusable Skill to a team: organization-provisioned Skills that reach every member at once, plugin-bundled Skills assigned to specific groups with versioned updates and rollback, Claude Code project Skills that live in a repository under .claude/skills/, and API Skills invoked programmatically by a partner's own products. Server-managed Claude Code settings are a separate configuration channel delivered on an hourly polling cycle - not a Skills distribution path.
Reuse at team scale means choosing how a Skill travels
A Skill packages a repeatable procedure into a versioned, reusable unit. Once you have one, the team-scale question is how to get it to everyone who should have it. The Claude Certified Architect - Professional (CCAR-P) exam treats this as a remember-level skill: you need the four distribution mechanisms firmly in memory, along with the one thing that looks like a fifth mechanism but is not. Distributing a workflow as a Skill is how a good local practice becomes a team standard - the procedure travels as one governed artifact instead of as undocumented know-how.
Each mechanism differs on two axes that matter: who can reach the Skill, and how much control you retain over it. Getting the four straight is what lets you pick correctly later, so this knowledge point is about knowing them cold before you have to choose between them.
- Skills distribution mechanisms
- The four channels for delivering a reusable Skill to a team: organization-provisioned (available to every member at once), plugin-bundled (assigned to specific groups with versioning and rollback), Claude Code project Skill (a filesystem artifact in .claude/skills/ that versions with a repository), and API Skill (invoked programmatically by a partner’s own products). Server-managed Claude Code settings are a separate configuration channel, not a distribution mechanism.
The four mechanisms
Organization-provisioned Skills are uploaded under Organization settings, Skills, and become available to every member of the organization at once. This is the simplest path when a capability genuinely should reach all members. It does not, on its own, give you group targeting or a rollback path - it is broad availability, not governed distribution.
Plugin-bundled Skills wrap one or more Skills into a plugin that you assign to a specific group. Only that group's members can access the bundled Skills. Plugins are also where governed distribution lives: install preferences, group targeting, version-controlled updates from a connected repository, and rollback. When a Skill should reach only select members, or needs to be centrally revocable, this is the mechanism.
Claude Code project Skills are filesystem artifacts that live in the project repository under .claude/skills/. They version with the repository itself and are scoped to the projects that carry them, which makes them the natural home for a coding convention or tool set that belongs to a specific codebase.
API Skills are invoked programmatically by a partner's own products through the API, rather than surfaced to a human through a chat interface. When the consumer of the capability is another product calling it in code, this is the path.
The thing that is not a distribution mechanism
The trap this knowledge point exists to inoculate against is server-managed Claude Code settings. Centrally managed configuration is delivered from Anthropic's servers when users authenticate and refreshes on an hourly polling cycle. It is a powerful way to enforce configuration across a team - but it configures settings, it does not distribute Skills. On the exam, "centrally managed configuration" appears as a plausible-looking distribution answer specifically so you can identify it as the wrong category. A settings channel and a Skills distribution channel are different things.
What the CCAR-P exam trips candidates on
Two traps recur. The first is selecting "centrally managed configuration" as a Skill distribution answer. It configures settings on an hourly polling cycle; it does not distribute Skills. If an option describes managed settings and the question asks how to distribute a Skill, it is a distractor by construction.
The second is recommending an organization-provisioned Skill when the scenario requires the capability to reach only a specific group. Org-provisioned means everyone. If the requirement is a subset - one department, one team - the correct mechanism is a plugin assigned to that group, because org-wide provisioning cannot scope to a subset.
Worked example
An Architect lists the options for sharing a reusable Skill with a team and a colleague suggests four: an organization-provisioned Skill, a plugin distributed to a group, a Claude Code project Skill, and enabling server-managed settings that push the Skill to everyone. Which of these does not belong on the list, and what are the real four mechanisms?
The one that does not belong is "server-managed settings that push the Skill to everyone." Server-managed Claude Code settings are a configuration channel - they deliver settings from Anthropic's servers at authentication and refresh on an hourly polling cycle. They enforce configuration, but they do not distribute Skills. Framing them as a way to push a Skill conflates two different systems, and that conflation is exactly the mistake the exam probes.
The real four are organization-provisioned Skills, which reach every member at once; plugin-bundled Skills, which target specific groups and carry versioning and rollback; Claude Code project Skills, which live in .claude/skills/ and version with a repository; and API Skills, which partner products call programmatically. The colleague's list had three genuine mechanisms and one impostor. The corrected list swaps managed settings out for API Skills, which is the fourth genuine mechanism and the one they missed.
The reason to keep these straight is that the next decision - which mechanism to actually use - depends on knowing the full, correct set. Miss API Skills and you have no answer when the consumer is another product; treat managed settings as a distribution path and you will reach for it when the real requirement was group-scoped, revocable delivery, which only a plugin provides.
Common misreadings to avoid
Misconception
Centrally managed configuration is a fifth way to distribute a Skill to a team.
What's actually true
Misconception
An organization-provisioned Skill can be scoped to reach only one department.
What's actually true
How this shows up on the exam
Domain 7 questions on this knowledge point list distribution options and ask which fits, or ask you to identify the option that is not a distribution mechanism at all. The reliable frame is the four genuine mechanisms - org-provisioned, plugin, project Skill, API Skill - plus the recognition that managed settings are a separate configuration channel.
This knowledge point builds on the shared team environment baseline, since Skills are the reusable assets that ride on top of that baseline. It sets up plugin governance for shared assets, which zooms in on what a plugin adds over a flat bundle, and choosing the right skill distribution mechanism, where you weigh these four against a described scenario. Know the four cold and the analysis becomes straightforward.
Which of the following is NOT a way to distribute a reusable Skill to a team?
People also ask
What are the ways to distribute a Skill to a team?
Is centrally managed configuration a way to distribute Skills?
What is the difference between a project Skill and an API Skill?
Watch and learn
Official Anthropic Academy lessons first, then hand-picked walkthroughs. Videos load only when you press play.
No videos curated for this concept yet
We are still curating the best official and community videos for this topic.
Official prep for this domain
Anthropic's own free prep module for this part of the syllabus, on the official prep course. Free with an Anthropic Academy sign-in.
References & primary sources
Master this concept with Archie
Practice it inside an adaptive study session. Archie, your Socratic AI tutor, tracks your mastery with Bayesian Knowledge Tracing and schedules the perfect next review.