Claude Code for Teams: Rules, Skills, Hooks, Plugins

What's inside

1. Level 1: The Extension Map: Which Surface Owns Which RuleFree
   Survey every Claude Code customisation surface (CLAUDE.md, rules, settings.json, skills, commands, subagents, hooks, MCP servers and plugins) and learn the key split between guidance Claude reads as context and configuration the client enforces, in a course that is independent of Anthropic.
2. Level 2: Layered CLAUDE.md: Scopes, Load Order and ImportsFree
   Build a CLAUDE.md hierarchy across organisation, project, user and local scopes, and understand what loads at session start, what loads on demand from subdirectories, and how the layers combine.
3. Level 3: Path-Specific Rules in .claude/rules/
   Split instructions into topic files under .claude/rules/ and use the paths frontmatter so a rule only enters context when Claude works on matching files.
4. Level 4: Team settings.json: Permissions and Precedence
   Write a shared .claude/settings.json with allow, ask and deny permission rules, and learn the precedence order from managed settings through command line, project local and shared project down to user settings.
5. Level 5: Your First Skill: SKILL.md and Its Frontmatter
   Create a project skill as a folder containing SKILL.md, and learn what the main frontmatter fields do and how a skill's body enters and stays in the conversation.
6. Level 6: The Description Is the Trigger: Testing Whether It Fires
   Tune the description and when_to_use text that Claude reads to decide whether to load a skill, and build a repeatable test to check it fires when it should and stays quiet when it should not.
7. Level 7: Progressive Disclosure: Scripts and Reference Files
   Keep SKILL.md short by moving detail into reference files and executable scripts that load or run only when needed, so a rich skill costs little context until used.
8. Level 8: Slash Commands with Arguments
   Build user-invoked commands as skills, pass arguments into them, and control whether the user, Claude or both may invoke each one, with legacy .claude/commands/ files as the older single-file format.
9. Level 9: Sharing and Troubleshooting Skills
   Distribute skills at project, personal and organisation-managed level, understand which copy wins on a name clash, see in outline how skills work in the Claude apps and the API, and work through a checklist when a skill misbehaves.
10. Level 10: Subagents: Context Isolation and Definition Files
    Define a subagent as a Markdown file with frontmatter in .claude/agents/, and understand what its fresh context contains and what it cannot see from the main conversation.
11. Level 11: Tools, Model and the Brief: Shaping a Subagent
    Restrict a subagent's tools, pick its model and permission mode, and write a delegation brief that gets back a result the main session can act on.
12. Level 12: Parallel and Background Agents, and When Not to Use One
    Run several subagents at once or in the background, understand the reduced tool set and permission flow for background work, and recognise tasks where delegation costs more than it saves.
13. Level 13: Hook Fundamentals: Events, Matchers and Handlers
    Wire a first hook in settings.json by choosing a lifecycle event, a matcher and a handler type, and read the JSON a hook receives on stdin.
14. Level 14: Blocking, Rewriting and Redacting Tool Calls
    Turn a must-always-hold rule into a deterministic guardrail using exit codes and JSON decisions, covering denying a tool call, rewriting its input before it runs and scrubbing output afterwards.
15. Level 15: Securing the Hooks Themselves
    Treat hooks as code that runs with your own privileges: review where they come from, harden the scripts, and limit which hooks may run across a team.
16. Level 16: MCP Servers: Scopes and Authorisation
    Connect MCP servers to Claude Code at local, project and user scope, share a team server through .mcp.json, and authorise it without committing credentials.
17. Level 17: Tool Bloat and Context Cost
    Measure what every connected server, skill listing and memory file costs in context, and trim capabilities so Claude picks the right tool more reliably.
18. Level 18: Packaging a Plugin
    Bundle skills, commands, agents, hooks and MCP servers into one plugin directory with a plugin.json manifest, and validate and test it locally.
19. Level 19: Marketplaces: Versions, Dependencies and Trust
    Publish plugins through a marketplace.json catalogue in a git repository, control versions and dependencies, and set the organisational controls that decide which marketplaces and plugins a team may install.
20. Level 20: Final Challenge: Design and Govern a Team Setup Exam
    Review the whole course by designing a complete team configuration for a realistic scenario, choosing the right mechanism for each need, justifying its context cost and defining how the shared configuration is owned and changed.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →