Claude Code: The Features You're Not Using

⚡

Start every project with /init

Running /init at the start of a new project lets Claude scan your entire codebase before you write a single prompt.

What /init does for you

Scans the entire codebase structure
Understands architecture & key files
Detects coding patterns and conventions
Creates a CLAUDE.md summary file

CLAUDE.md — Your Project's Persistent Brain

Think of CLAUDE.md as a README written for Claude, not humans. It's included automatically in every request, meaning Claude always starts with full project context — no re-explaining needed.

Guides Claude on architecture, file structure & key commands

Stores Custom instructions and project-level preferences

Auto-loads Included in every conversation, zero manual effort

📁

CLAUDE.md

Project-level. Committed to your repo. Shared with the whole team.

🔒

CLAUDE.local.md

Personal preferences. Not committed. Not shared. Yours alone.

🌍

~/.claude/CLAUDE.md

Global rules. Applies to every project you open. Set once, always active.

Writing Custom Instructions with #

Prefix any message with # to enter memory mode. Claude writes it straight into CLAUDE.md.

        # Enter memory mode

        # Use comments sparingly.

        # Only comment complex logic.

        → Claude auto-merges into CLAUDE.md

Referencing Files with @

Use @filename to pull a file into the conversation. Works inline in prompts or inside CLAUDE.md for persistent context.

        # In a prompt

        How does auth work? @auth.ts

        # In CLAUDE.md (persistent)

        @prisma/schema.prisma

⛔

Esc — Interrupt & Redirect

Press Esc once to stop a response mid-stream. Redirect Claude immediately to where you actually want it to go. Prevents deep, off-track rabbit holes.

⏪

Esc Esc — Rewind the Conversation

Press Esc twice to strip away distracting history while keeping valuable context. Ideal when a conversation has gone sideways.

🗂 Context Management Quick Reference

Esc Interrupt the current response — redirect immediately

Esc Esc Rewind — remove distracting messages, keep useful context

/compact Summarise conversation — retains knowledge, reduces noise

/clear Full reset — removes all conversation history entirely

Best Use Cases for Context Controls

Long-running development sessions spanning hours or days
Complex multi-module projects with lots of file context
Correcting repeated mistakes without resetting everything
Context-heavy AI workflows where relevance degrades over time

Creating a Custom Command

Create .claude/commands/ folder in your project
Add a markdown file (e.g. audit.md)
The filename becomes the command name (/audit)
Restart Claude Code to activate

Commands with Arguments

Use $ARGUMENTS as a placeholder to accept dynamic input — file paths, text, flags. Makes commands reusable across any context.

        # .claude/commands/review.md

        Review the following file for

        code quality issues:

        $ARGUMENTS

        # Usage:

        /review src/auth.ts

What are Hooks?

Hooks let you intercept Claude's tool calls — automatically running your own commands before or after Claude acts. Format, test, validate, log, or block — all without lifting a finger.

🔀

PreToolUse vs PostToolUse

Hook	When	Can Block?
PreToolUse	Before tool runs	✅ Yes
PostToolUse	After tool runs	❌ No

Building a Hook — 4 Steps

Choose Pre or Post hook timing
Select which tool types to monitor
Write the command to process the call
Return feedback or block the operation

All Available Hook Types

PreToolUse Before tool execution — can block

PostToolUse After tool execution — cannot block

Notification When Claude sends a notification

Stop When Claude finishes responding

SubagentStop When a subagent task finishes

PreCompact Before /compact runs

UserPromptSubmit Before Claude processes your prompt

SessionStart / SessionEnd Session lifecycle events

Practical Hook Applications

Auto-format code after every edit
Run tests on file changes automatically
Block access to sensitive files (.env, keys, .git)
Enforce coding standards project-wide
Track and log all Claude tool usage

Hook Configuration

Hooks are defined in .claude/settings.json. You can scope them globally, per-project, or locally.

        // .claude/settings.json

        {

          "PreToolUse": [...],

          "PostToolUse": [...]

        }

Understanding Hook Payload

Hooks receive JSON via stdin. Structure varies by type — use a logging helper to inspect the real payload.

        {

          "session_id": "...",

          "hook_event_name": "...",

          "tool_name": "...",

          "tool_input": {...},

          "tool_response": {...}

        }

🔐 Security Best Practices

Validate & sanitize all hook inputs
Always use absolute file paths
Block path traversal attempts explicitly
Avoid touching sensitive files (.env, key files, .git)

🐛 Debugging Hooks

Hook input structure can be unpredictable. Log the raw stdin payload first — it saves hours of guesswork.

        # Helper hook to log payload

        cat - | jq . > post-log.json

        # Then inspect:

        cat post-log.json

Claude Code:
The Features You're Not Using

Initialisation & Project Memory

Start every project with /init

What /init does for you

CLAUDE.md — Your Project's Persistent Brain

CLAUDE.md

CLAUDE.local.md

~/.claude/CLAUDE.md

Writing Custom Instructions with #

Referencing Files with @

Context Control

Esc — Interrupt & Redirect

Esc Esc — Rewind the Conversation

🗂 Context Management Quick Reference

Best Use Cases for Context Controls

Custom Commands

Creating a Custom Command

Commands with Arguments

Hooks

What are Hooks?

PreToolUse vs PostToolUse

Building a Hook — 4 Steps

All Available Hook Types

Practical Hook Applications

Hook Configuration

Understanding Hook Payload

🔐 Security Best Practices

🐛 Debugging Hooks

Claude Code:The Features You're Not Using

Initialisation & Project Memory

Start every project with /init

What /init does for you

CLAUDE.md — Your Project's Persistent Brain

CLAUDE.md

CLAUDE.local.md

~/.claude/CLAUDE.md

Writing Custom Instructions with #

Referencing Files with @

Context Control

Esc — Interrupt & Redirect

Esc Esc — Rewind the Conversation

🗂 Context Management Quick Reference

Best Use Cases for Context Controls

Custom Commands

Creating a Custom Command

Commands with Arguments

Hooks

What are Hooks?

PreToolUse vs PostToolUse

Building a Hook — 4 Steps

All Available Hook Types

Practical Hook Applications

Hook Configuration

Understanding Hook Payload

🔐 Security Best Practices

🐛 Debugging Hooks

Claude Code:
The Features You're Not Using