Install & Use

RBA Skills ships as a Claude Code plugin marketplace. The catalog you see on this site is the same set of bundles you install in your terminal, your IDE, your desktop app, or claude.ai/code. One command connects you to every plugin RBA has published; another installs a specific plugin into your session.

This page covers the daily use loop — install, invoke, update. For the canonical install instructions with copyable commands and screenshots, see Plugins → Install.

Prerequisites

You need Claude Code. It’s available everywhere Claude is:

Desktop apps for Mac and Windows — claude.com/download
Web app — claude.ai/code
Terminal CLI — code.claude.com/docs/en/setup
IDE extensions for VS Code and JetBrains

The slash commands below work identically on every surface.

1. Add the marketplace (one-time)

Open any Claude Code session and run:

claude plugin marketplace add rbaconsulting/rba-skills-plugin

This registers the RBA Skills marketplace with your Claude Code install. You only need to do this once per machine.

2. Install a plugin

Pick a plugin from the catalog — for example, research or accessibility. Inside Claude Code:

claude plugin install <plugin-name>@rba-ai-hub

You’ll be prompted to choose an install scope:

User — available across every project on your machine
Project — committed alongside your repo and shared with collaborators
Local — only this checkout

If you prefer a UI to slash commands, run /plugin to open the interactive plugin manager.

3. Invoke what the plugin ships

A plugin can ship up to seven kinds of artifact. Here’s how each one shows up in a session.

Skills are methodologies Claude follows during a conversation. Two ways to invoke:

Explicit: type /<skill-name> followed by your task. Example: /research compare Sitecore XM Cloud vs Optimizely SaaS CMS for headless delivery.
Implicit: describe your task in plain English. Claude reads the skill’s description and triggers it when the context matches.

A well-written SKILL.md works either way. Use explicit invocation when you want the methodology now; use implicit when you want Claude to pick the right pattern for you.

Agents are specialized subagents — a separate Claude instance with its own system prompt, tool allowlist, and context window. Use one when the work is scoped, isolated, and benefits from a specialist viewpoint.

Ask Claude to “use the <agent-name> agent to …” and it will spawn the subagent and report back the result. The subagent’s working context is discarded after the task, keeping your main thread clean.

A slash command is the most direct invocation — type /<command-name> and Claude runs it. New work usually ships as a skill instead (skills can be auto-discovered; commands have to be typed). Slash commands are kept for cases where you want strictly user-triggered behavior — a checklist, a one-shot formatter, an explicit gate.

Hooks run automatically — you don’t invoke them. Installing a plugin that ships hooks wires them into the Claude Code lifecycle events declared by the plugin (PreToolUse, PostToolUse, Stop, etc.).

Hooks are how RBA enforces guarantees that prompts can’t — formatter integration, governance gates, audit trails. They run deterministically every time the matching event fires. Uninstall the plugin and they unwire cleanly.

Resources are non-executable assets — project skeletons, templates, reference architectures, runbooks. After installing the plugin, the files are on disk under the plugin’s install path. Read them, copy from them, or point Claude at them with @<path>.

Using RBA Skills without Claude Code

You don’t have to install anything to get value from the catalog. Every artifact has a page on this site:

Browse by tier: Plugins, Skills, Agents, MCP, Hooks, Commands, Resources, or search everything.
Read the methodology: Every skill’s page renders its full SKILL.md. The methodology travels — you can adapt it for ChatGPT, Codex, an internal model, or just paper-and-pen.
Copy and adapt: SKILL.md is a portable artifact. Treat the on-site copy as the canonical version and lift what you need.

This matters for two reasons. First, not everyone at RBA is in a Claude Code workflow today — a content strategist or a BA might be using web Claude or another tool entirely. Second, the methods are the durable thing. The wrapper around them is a vendor choice that may evolve.

Updating and removing plugins

Inside any Claude Code session:

List installed plugins — /plugin list
Update a plugin — /plugin update <plugin-name>@rba-ai-hub
Remove a plugin — /plugin remove <plugin-name>
Browse interactively — /plugin

Updates pull the latest plugin version from the marketplace and re-wire any hooks, MCP servers, or commands the plugin ships. Removing a plugin tears those down cleanly.

Tips for getting good results

Be specific. “Audit accessibility” is generic. “Audit WCAG 2.1 AA compliance for the checkout flow in our Next.js storefront, focusing on keyboard navigation and screen reader announcements” gets a useful answer.
Provide real context. Paste actual code, real meeting notes, real requirements. The methodology in a skill is only as good as the inputs you give it.
Iterate, don’t restart. If the first output is 80% right, ask for refinements. Don’t start over — Claude remembers the conversation.
Combine plugins. Run /research to gather context, then a platform plugin’s skill to act on it. The catalog is designed to compose.
Save good outputs. When a skill produces something great, keep the output as a reference. Your filled-in prompts are reusable too.

Next steps

Plugins catalog — Browse every installable plugin in the catalog.
Skills catalog — See every methodology across every plugin.
Contributing — Ship a new plugin or add a skill to an existing one.
AI Governance — Policies and guardrails for AI use at RBA.