🏴‍☠️ Pirated From Another Group - CC Features

I use about 10 of these 31 features. In some cases, like making Claude.md files, and hooks, I have CC write them.

Some of the features are over my head, but all I’d do to figured out is ask Claude to tell me about it.

For Claude Code users, here’s a thread I followed quite regularly since December thought it’s a great resource. Had Grok give a breakdown of all 31 features.

https://x.com/adocomplete/status/2006802341228655083?s=20

31 Days of Claude Code Features Breakdown

1 ! Prefix for Bash Commands

The ! prefix allows you to execute bash commands instantly and inject the output directly into the conversation context without model processing, delays, or token waste. Example: Type “!git status” to run it and add the results to Claude’s knowledge. Useful for quick checks like running tests or commands without unnecessary back-and-forth.

2 Double Esc to Rewind

Press Esc twice to rewind the conversation, code changes, or both to a previous checkpoint. This lets you experiment with “what if” scenarios and revert if things go wrong, preserving momentum without losing progress.

3 Extended Thinking in API

In the Claude API, add thinking: { type: “enabled”, budget_tokens: 5000 } to enable Claude to show step-by-step reasoning in thinking blocks before responding. This provides transparency into Claude’s thought process for complex queries.

4 Session Management

Use claude --continue to resume the last session instantly or claude --resume to pick from past sessions via a picker. This preserves context even if your terminal closes or device restarts, maintaining workflow continuity.

5 # Prefix for Memory Updates (Removed)

Originally, starting a message with # (e.g., “# Always use bun instead of npm”) allowed quick updates to Claude’s memory files (ClaudeMD). This feature has been removed; now, you can directly ask Claude to remember details or update files.

6 Vim Mode

Type /vim to enable full Vim-style editing in the prompt area, with commands like h/j/k/l for navigation, ciw to change a word, dd to delete a line, or w/b to jump words. Ideal for power users who want keyboard-only prompt editing at high speed.

7 Headless Mode

Add -p flag for non-interactive use in scripts, CI/CD, or pipes (e.g., claude -p “Fix lint errors” or git diff | claude -p “Explain changes”). Enables integrating Claude into automated pipelines without a UI.

8 /context Command

Type /context to inspect what’s consuming your token window, including the system prompt, MCP servers, memory files, and more. Provides visibility into context usage for optimization.

9 Claude Code Remote

Use & after a prompt to send tasks to Claude Code on the web (runs in background). Then, claude --teleport session_id pulls the session to your local terminal. Allows seamless switching between local and remote environments.

10 /stats Command

Type /stats to view usage statistics like a contribution graph, showing how you’re using Claude Code. A fun way to track productivity, similar to GitHub streaks but for AI interactions.

11 Named Sessions

Use /rename session-name to name the current session, then /resume session-name or claude --resume session-name to reload it. Makes managing multiple workflows easier, like naming git branches.

12 Ultrathink Mode

Include “ultrathink” in your prompt to allocate up to 31,999 thinking tokens for deep reasoning. (Other modes like “think” or “think hard” have been removed.) Great for complex problems requiring extensive internal deliberation.

13 Prompt Stashing

Press Ctrl+S to stash (save) a draft prompt, send something else, and have it auto-restore later. Prevents losing your train of thought without needing external notepads.

14 /statusline Customization

Use /statusline to customize the status bar with relevant info like git status, model details, tokens used, weather, or custom data. Configurable in settings.json for personalized at-a-glance insights.

15 YOLO Mode (–dangerously-skip-permissions)

Launch with claude --dangerously-skip-permissions to auto-approve all actions without prompts. Speeds up workflows but use cautiously due to potential risks (hence the “danger” warning).

16 Hooks

Hooks are shell commands that run at specific lifecycle events (e.g., before/after tool calls). Configure via /hooks or settings.json to add determinism, like blocking dangerous commands or sending notifications.

17 @ Mentions

Use @ to quickly add files/directories to context, list contents, or enable/disable/invoke MCP servers/subagents (e.g., @file.ts or @dir/). Fastest way to provide Claude with targeted context.

18 Prompt Suggestions

After a task, Claude suggests grayed-out follow-up prompts. Press Tab to accept and edit, or Enter to run immediately. Automates workflow continuation by predicting next steps.

19 Claude Code + Chrome Integration

Type /chrome to let Claude interact with your browser: navigate pages, click buttons, read errors, inspect DOM. Enables end-to-end tasks like fixing bugs and verifying in-browser.

20 Agent Skills

Skills are folders with instructions, scripts, and resources for specialized tasks. Packaged as an open standard, they’re reusable across agents, teaching Claude niche abilities once for repeated use.

21 Plugins

Use /plugin install name to install bundled workflows (commands, agents, skills, hooks, MCP) from a marketplace. Simplifies sharing and discovering complete setups in one command.

22 Ctrl + R for Reverse Search

Press Ctrl+R to search prompt history backward, cycle matches with additional presses, Enter to run, or Tab to edit. Quick way to recall and reuse past prompts without retyping.

23 /init Command

Type /init to have Claude scan your codebase and generate a CLAUDEmd file with onboarding docs (build commands, tests, key dirs, conventions). Auto-onboards Claude to new projects.

24 /export Command

Type /export to dump the entire conversation (prompts, responses, tool calls) to markdown. Useful for documentation, training, or reviewing past work.

25 Subagents

Subagents are parallel workers with their own 200k context windows for specialized tasks, merging outputs back to the main agent. Enables delegation for complex, multi-part jobs.

26 Commands

Save prompts as reusable slash commands in markdown files, supporting arguments (e.g., /explain src/auth.ts). Reduces repetition for common tasks.

27 Sandbox Mode

Type /sandbox to set boundaries where Claude can operate freely without per-action permissions. Combines speed of YOLO mode with security controls.

28 /usage Command

Type /usage to check rate limits, or /extra-usage to request more. Helps monitor and expand capacity as needed.

29 Plan Mode

Press Shift+Tab twice to enter Plan mode, where Claude explores/analyzes the codebase without edits. Encourages thinking before acting; recommended for 90% of tasks.

30 Language Server Protocol (LSP)

LSP integration provides IDE-like features: instant diagnostics (errors/warnings), navigation (go to definition, find references), and type/docs hover. Brings advanced code intelligence to the terminal.

31 Claude Agent SDK

The SDK exposes the agent loop, tools, and context management from Claude Code, allowing you to build custom agents in ~10 lines of code. Opens up creating Claude-like experiences for any application.

2 comments

Claude Code Pirates

skool.com/claude-code-pirates-8106

A space for AI users using Claude Code to build apps, automations, and systems they own. No hype.

Ashish Builds Academy – Lite

Vibe Coding Academy

AI Automations by Jack

Bring people together around your passion and get paid.