Video summary
Every Claude Code Command in 13 minutes! (10x your productivity)
Main summary
Key takeaways
Core idea / product framing
The video ranks every “Claude Code” slash command into 10 productivity tiers, emphasizing which commands are most practical and how to use them across workflows—from setup to multi-agent automation.
Claude Code commands are easy to access: type / to open a command list. Commands are positioned as tools for:
- Initial setup (persistent project context, authentication)
- Session control (resetting context, switching models, managing permissions)
- Code operations + safety (diffs, rollback, merge flows, vulnerability scanning)
- Automation and power-user orchestration (batching, loops, MCP tools)
- IDE integrations and output (exporting, GitHub/Slack integrations, terminal/VS Code–style tooling)
Tier 1 — “Non-negotiables” (setup & baseline config)
/login: Authenticates with an Anthropic account (name + password)./doctor: Diagnoses the installation and flags issues before starting./init(initialization): Createsclaude.mdto establish persistent context/memory (tools, languages, forbidden folders, code style, answer detail level, and personal/business/audience info). Updates over time./config//settings: Opens settings UI:- Theme (light/dark)
- Default model
- Output preferences (including memory-cleaning behavior and explanation level)
/help: Shows available commands and basic usage guidance.
Tier 2 — Session control & cost/performance management
These focus on keeping the session healthy, reducing token waste, and controlling what Claude can do.
/clear,/reset,/new,/again: Clear conversation history and free context for a fresh start (use when responses degrade, hallucinations increase, performance slows, etc.)./model(important for cost): Switch AI models mid-session immediately.- Haiku: quick/cheap for easy tasks
- Sonnet: “sweet spot” for most tasks (primary usage by the speaker)
- Opus: highest token cost, best for complex/multi-agent work
/cost: Shows tokens used and current cost./usage: Displays plan limits and rate-limit status (how close you are to being blocked)./permissions(allowed tools control):- Prevent writing to certain folders
- Require permission to run terminal commands
- Restrict file types Claude can change
- Protect secrets/passwords
/compact: Compresses conversation history to free context without fully wiping it (keeps important context)./context: Visualizes context window fullness and warns before it becomes problematic./plan: Enables “planning mode” so Claude thinks through the problem before executing—reduces wasted tokens and wrong actions.
Tier 3 — Code changes, reviews, branching, and memory
/diff: Interactive viewer showing uncommitted changes and what Claude changed per turn./rewind(checkpoint): Roll conversation + codebase back to a previous point if things go wrong./fork: Branch the conversation to try multiple approaches; then merge best changes back together (no work lost)./memory: Edits theclaude.mdmemory rules established at init; good for fixing recurring preferences/errors./security review: Scans pending changes for vulnerabilities before shipping (detects things like leaked passwords/login issues). Built-in and presented as underused./simplify: Spawns three parallel review agents to:- find code quality issues in recently changed files
- fix them
- let you specify focus (e.g., speed vs memory)
/tasks: Lists/manages background tasks Claude is running./btw(new, emphasized): Ask a side question without polluting the main conversation context while Claude keeps working./debug: Troubleshoots the current session by reading session logs; can focus on a specific issue.
Tier 4 — Power-user automation (“missions” and orchestration)
/MCP: Connect Claude Code to third-party platforms via MCP servers (examples mentioned: Notion, Slack, Google Drive)./batch: Splits a large job into smaller parallel tasks; opens separate GitHub pull requests per subtask for review before merging back./loop: Repeats a prompt on an interval (e.g., check deploy status, babysit PRs, schedule check-ins; retry tests until passing; poll APIs for changes)./Claude API: Loads API reference for the current language; activates when Anthropic SDK imports are detected./plugin: Installs/removes/manages plugins that extend Claude Code./agents: Manages multi-agent configurations./hooks: Automated actions triggered by tool events (e.g., run code checker on file save, notify when Claude finishes, run tests after changes)./reload plugins: Reload active plugins without restarting./skills: Lists installed skills Claude can use for specialized tasks (speaker notes 60,000+ skills exist in the ecosystem).
Tier 5 — Workflow & output (what to do with results)
/export: Saves the current conversation as a plain text file./copy: Copies the last response; includes a picker for individual code blocks./PR comments: Pulls and displays GitHub PR comments in the current session./install GitHub app: Sets up Claude GitHub actions integration for a repository./install Slack app: Installs via OAuth so Claude can act inside Slack.
Tier 6 — IDE/environment integration
/IDE: Manages IDE integrations and shows connection status./add dir: Adds an additional working directory to the session./desktop//app: Moves the current session to the Claude Code desktop app.- Terminal setup / key bindings commands:
- Terminal key bindings (e.g., shift+enter behavior mentioned)
/key bindings: Opens shortcuts config for customization
/sandbox: Toggles sandbox mode so Claude runs in isolation and doesn’t affect other systems.
Tier 7 — Personalization & UI “cosmetics”
/theme: Color theme including light/dark/color-blind accessible./color: Sets prompt bar color for the session./status line: Configures what Claude Code shows in the terminal status line./vim: Toggles editing mode between Vim and normal.
Tier 8 — Informational/analytics
/status: Shows version, model, account, connectivity./stats: Visual breakdown of usage, session history, streaks./insights: Report analyzing usage patterns and friction points./release notes: Full changelog (latest first)./effort: Sets model effort level low → max (trades speed for depth)./fast: Toggles fast mode to get quicker responses and reduce token usage/cost./privacy settings: View/update privacy settings (available only on Pro and Max plans).
Tier 9 — Rare/situational & account/advanced controls
Mentioned as “when nothing else works” / infrequent:
/rename: Rename the session (shown on prompt bar)./resume: Reopen previous session by ID/name/pair./RC(remote control): Control current session remotely via Claude AI (described like an “AI employee”)./remote ENV: Default remote environment for web-based sessions./feedback//bug: Submit bug reports to Anthropic./extra usage: Add more capacity after hitting rate limits (costs money)./passes: Eligible accounts can gift a free week./Chrome: Configure settings for a browser agent (Claude actions in the browser)./mobile: Shows QR code for downloading iOS/Android app./upgrade: Upgrade plan tier./log out: Sign out of Anthropic account./exit//quit: Close Claude Code session.
Tier 10 — Deprecated / special
/Review: Deprecated; points to installing a code review plugin instead./stickers: A humorous aside—orders Claude Code stickers (speaker calls it arguably the “best command”).
Main speaker/source
- Duncan Rogoff (creator; former art director for Apple/PlayStation/Nissan; runs an AI agency and community “Build Room”).