Topic catalogue

For all engineers

• Progressive disclosure: best practices for CLAUDE.md files and rules Recommended
  The CLAUDE.md hierarchy, path-scoped rules, and keeping instructions short enough to be followed.
• Context windows: what fills them and how to keep them clear Recommended
  Reading /context, compacting, and knowing when to start fresh with /clear.
• How Claude Code works, and choosing a permission mode
  The agent loop, what each permission mode allows, and when auto mode fits.
• Plan first, then review and rewind
  Plan mode, reading diffs properly, and recovering with checkpoints and /rewind.
• Git and pull requests with Claude Code
  Commits, branches, pull requests and working in worktrees.
• Coming from Copilot or Cursor
  What maps across from other AI coding tools, and what works differently.
• Skills and slash commands
  Writing a skill your team will reuse, and checking it with /skill-doctor.
• Hooks: guardrails that always run
  Why CLAUDE.md is advice and hooks are enforcement, with examples of each.
• MCP servers: connecting your tools and data
  What MCP gives Claude Code, and choosing which servers to connect.
• Sub-agents, agent teams and parallel work
  Background sub-agents, worktrees and /fork, and when parallel work pays off.
• Automating Claude Code: headless, CI, routines and /loop
  Running Claude Code on a schedule or a trigger, without anyone at the keyboard.
• Code review with Claude Code
  Using /code-review locally and running review in CI.
• Plugins and mods: sharing and customising Claude Code
  Plugins, marketplaces, and mods such as panels, status lines and commands.
• Asking well: prompts, context and course-correcting
  Stating outcomes, pointing at the right files, and interrupting early.
• Exploring and debugging an unfamiliar codebase
  Asking before changing, and tracing a bug back to its cause.
• Tests as the feedback loop
  Writing tests first, then letting Claude run them and iterate.
• Claude Code in your IDE, desktop and browser
  VS Code, JetBrains, the desktop app and the web, and what each suits.
• Large refactors and codebase migrations
  Splitting the work, running it in batches, and keeping it reviewable.
• Choosing models and effort Recommended
  Using /model, effort levels and fast mode, and knowing when to spend more.
• Driving the browser to verify UI changes
  Claude in Chrome, and reading console and network output.
• Security reviews with Claude Code
  What /security-review does, and what it does not catch.

For Staff+ engineers: guidelines and standards

Organisation only These topics are in the Organisation package's dedicated session.

• Layering CLAUDE.md and rules across repositories and teams
  Organisation, repository and path-scoped instructions that stay consistent.
• Standards that enforce themselves: hooks and checks
  Moving your rules from advice to enforcement.
• Shared skills, plugins and mods: ownership, review and versioning
  Running a team marketplace, and validating what goes into it.
• Reviewing AI-written code, defining done and measuring impact
  Review standards, quality gates, analytics and pull request attribution.

For administrators: settings and policies

Organisation only These topics are in the Organisation package's dedicated session.

• Managed settings and permissions policy for every developer
  Managed settings files, MDM, server-managed settings and managed permission rules.
• Approving MCP servers, plugins and mods
  Allow and deny lists, and options only an administrator can set.
• Spend limits, analytics and telemetry
  Spend caps, the analytics dashboard and OpenTelemetry.
• Data, privacy and sandboxing
  How data is used, zero data retention, and sandboxing.