Claude Code Feature Reference: 31-Day Advent Compilation

Research Date: 2026-01-20
Publication Date: 2026-01-04
Author: Ado Kukic (Developer Relations, Anthropic)
Source URL: https://adocomplete.com/advent-of-claude-2025/

Reference URLs

• Original Tweet Announcement
• Advent of Claude Blog Post
• Claude Code Repository
• Agent Skills Standard
• Chrome Extension
• Author Twitter
• Author LinkedIn

Summary

This document compiles 31 Claude Code tips originally shared as daily posts on X/Twitter and LinkedIn throughout December 2025 by Ado Kukic, who leads developer relations at Anthropic. The compilation reorganizes the tips from beginner essentials to advanced patterns, providing a comprehensive reference for Claude Code’s feature set as of early 2026.

The tips span project initialization, keyboard shortcuts, session management, productivity features, thinking modes, permissions, automation, browser integration, and advanced agent extensibility patterns. Each section represents a coherent workflow category rather than the chronological order of original publication.

Getting Started

Day 1: /init — Project Onboarding

The /init command enables Claude to scan a codebase and generate a CLAUDE.md file containing:

• Build and test commands
• Key directories and their purposes
• Code conventions and patterns
• Architectural decisions

For larger projects, a .claude/rules/ directory supports modular, topic-specific instruction files. YAML frontmatter enables conditional rule application:

---
paths: src/api/**/*.ts
---
# API Development Rules
- All API endpoints must include input validation

Day 2: Memory Updates

Direct memory updates without manual file editing:

“Update CLAUDE.md: always use bun instead of npm in this project”

Claude appends the instruction to project memory. This replaces the previous # prefix syntax deprecated in Claude Code 2.0.70.

Day 3: @ Mentions — Context Injection

The @ symbol provides rapid context injection:

File suggestions are approximately 3× faster in git repositories and support fuzzy matching.

Essential Shortcuts

Day 4: The ! Prefix — Bash Execution

Direct shell command execution without token overhead:

! git status
! npm test
! ls -la src/

The ! prefix executes commands immediately, injecting output into context without model processing delay.

Day 5: Double Esc — Rewind

Pressing Esc twice returns to the last clean checkpoint. Options include:

• Rewind conversation only
• Rewind code changes only
• Rewind both

Note: Bash commands executed via ! cannot be undone.

Day 6: Ctrl+R — Reverse Search

Prompt history search functionality:

Compatible with slash commands.

Day 7: Ctrl+S — Prompt Stashing

Analogous to git stash for prompts. Ctrl+S saves the current draft, allowing interruption for other tasks. The draft auto-restores when the input field regains focus.

Day 8: Prompt Suggestions

After task completion, Claude may display grayed-out follow-up suggestions:

Configurable via /config.

Session Management

Day 9: Session Continuity

Recovery commands for interrupted sessions:

claude --continue    # Resume last conversation
claude --resume      # Display session picker

Session preservation duration configurable via cleanupPeriodDays setting (default: 30 days, 0 to disable).

Day 10: Named Sessions

Session naming for organized workflow management:

/rename api-migration           # Name current session
/resume api-migration           # Resume by name
claude --resume api-migration   # CLI resume by name

The /resume screen supports keyboard shortcuts: P for preview, R for rename.

Day 11: Claude Code Remote (Teleport)

Cross-platform session transfer:

# Start session on claude.ai/code
# Session runs in background

# Later, from terminal:
claude --teleport session_abc123

Transfers context, messages, and edits to local environment. Compatible with Claude mobile apps (iOS/Android) and Claude Desktop.

Day 12: /export — Session Export

Complete session export to markdown:

• All prompts sent
• All responses received
• All tool calls and outputs

Useful for documentation, training data, or audit trails.

Productivity Features

Day 13: /vim — Vim Mode

Full vim-style prompt editing:

Toggle with /vim.

Day 14: /statusline — Status Bar Configuration

Customizable terminal status bar elements:

• Git branch and status
• Current model
• Token usage
• Context window percentage
• Custom scripts

Day 15: /context — Token Inspection

Token consumption breakdown:

• System prompt size
• MCP server prompts
• Memory files (CLAUDE.md)
• Loaded skills and agents
• Conversation history

Day 16: /stats — Usage Statistics

Personal usage dashboard displaying:

• Model preferences
• Usage patterns over time
• Activity streaks
• Favorite features

Day 17: /usage — Rate Limit Monitoring

/usage         # View current usage with progress bars
/extra-usage   # Purchase additional capacity

Thinking & Planning

Day 18: Ultrathink

Extended thinking allocation via keyword:

> ultrathink: design a caching layer for our API

Allocates up to 32k tokens for internal reasoning before response generation. Supersedes the previous think / think harder / ultrathink graduated system. The keyword is ignored when MAX_THINKING_TOKENS environment variable is set.

Day 19: Plan Mode

Activated via Shift+Tab twice. In plan mode, Claude can:

• Read and search codebase
• Analyze architecture
• Explore dependencies
• Draft implementation plans

No edits occur until plan approval. Rejection includes feedback mechanism for iteration.

Day 20: Extended Thinking (API)

API-level thinking configuration:

thinking: { type: "enabled", budget_tokens: 5000 }

Exposes step-by-step reasoning in thinking blocks.

Permissions & Safety

Day 21: /sandbox — Permission Boundaries

Define operational boundaries once rather than per-request:

/sandbox

Supports wildcard syntax (e.g., mcp__server__*) for allowing entire MCP servers.

Day 22: YOLO Mode

Skip all permission prompts:

claude --dangerously-skip-permissions

Intended for isolated environments or trusted operations only. The flag name indicates risk level.

Day 23: Hooks

Lifecycle event handlers configured via /hooks or .claude/settings.json:

Use cases: block dangerous commands, send notifications, log actions, integrate external systems.

Automation & CI/CD

Day 24: Headless Mode

Non-interactive CLI execution:

claude -p "Fix the lint errors"
claude -p "List all the functions" | grep "async"
git diff | claude -p "Explain these changes"
echo "Review this PR" | claude -p --json

The -p flag outputs directly to stdout for pipeline integration.

Day 25: Custom Commands

Reusable prompts via markdown files:

/daily-standup              # Execute saved prompt
/explain $ARGUMENTS         # Parameterized command
/explain src/auth.ts        # Example invocation

Browser Integration

Day 26: Chrome Extension

Claude Code browser capabilities via claude.ai/chrome:

• Navigate pages
• Click buttons and fill forms
• Read console errors
• Inspect DOM
• Take screenshots

Enables single-prompt workflows like “fix the bug and verify it works.”

Advanced: Agents & Extensibility

Day 27: Subagents

Parallel agent spawning:

• Each subagent receives its own 200k context window
• Specialized task execution
• Parallel operation
• Output merges back to main agent
• Full MCP tool access

Subagents can run in background while main workflow continues.

Day 28: Agent Skills

Packaged instruction sets for specialized tasks:

• Folders containing instructions, scripts, and resources
• Portable across projects
• Open standard via agentskills.io
• Cross-tool compatibility

Examples: deployment processes, testing methodologies, documentation standards.

Day 29: Plugins

Bundled configurations:

/plugin install my-setup

Plugins package:

• Commands
• Agents
• Skills
• Hooks
• MCP servers

Marketplace discovery with search filtering available.

Day 30: LSP Integration

Language Server Protocol support provides:

• Instant diagnostics: Errors/warnings after each edit
• Code navigation: Go to definition, find references, hover information
• Language awareness: Type information and symbol documentation

Day 31: Claude Agent SDK

Programmatic agent construction:

import { query } from '@anthropic-ai/claude-agent-sdk';

for await (const msg of query({
  prompt: "Generate markdown API docs for all public functions in src/",
  options: {
    allowedTools: ["Read", "Write", "Glob"],
    permissionMode: "acceptEdits"
  }
})) {
  if (msg.type === 'result') console.log("Docs generated:", msg.result);
}

Exposes the same agent loop, tools, and context management used internally by Claude Code.

Quick Reference

Keyboard Shortcuts

Essential Commands

CLI Flags

Architecture Overview

References

1. Advent of Claude: 31 Days of Claude Code - Primary source, accessed 2026-01-20
2. Original Twitter Announcement - 2026-01-04
3. Claude Code GitHub Repository
4. Agent Skills Open Standard
5. Claude Chrome Extension

Note: The original 31 tips were published as daily posts on X/Twitter and LinkedIn throughout December 2025. The author’s compilation blog post reorganized the content thematically rather than chronologically. Individual tweet URLs are not included as they require authentication to access.