claude-hook-writer（Claude hook 写作）

Claude Hook Writer

Status: Production Ready Version: 2.0.0 (Optimized with progressive disclosure) Last Updated: 2025-12-17

Overview

Expert guidance for writing secure, reliable, and performant Claude Code hooks. This skill validates design decisions, enforces best practices, and prevents common pitfalls.

When to Use This Skill

• Designing a new Claude Code hook
• Reviewing existing hook code
• Debugging hook failures
• Optimizing slow hooks
• Securing hooks that handle sensitive data
• Publishing hooks as PRPM packages

Core Principles

1. Security is Non-Negotiable

Hooks execute automatically with user permissions and can read, modify, or delete any file the user can access.

ALWAYS validate and sanitize all input. Hooks receive JSON via stdin—never trust it blindly.

For complete security patterns: Load references/security-requirements.md when implementing validation or securing hooks.

2. Reliability Over Features

A hook that works 99% of the time is a broken hook. Edge cases (Unicode filenames, spaces in paths, missing tools) will happen.

Test with edge cases before deploying.

For reliability patterns: Load references/reliability-performance.md when handling errors or edge cases.

3. Performance Matters

Hooks block operations. A 5-second hook means Claude waits 5 seconds before continuing.

Keep hooks fast. Run heavy operations in background.

For performance optimization: Load references/reliability-performance.md when optimizing hook speed.

4. Fail Gracefully

Missing dependencies, malformed input, and disk errors will occur.

Handle errors explicitly. Log failures. Return meaningful exit codes.

Hook Design Checklist

Before writing code, answer these questions:

What Event Does This Hook Target?

• PreToolUse - Before tool execution (modify input, validate, block)
• PostToolUse - After tool completes (format, log, cleanup)
• UserPromptSubmit - Before user input processes (validate, enhance)
• SessionStart - When Claude Code starts (setup, env check)
• SessionEnd - When Claude Code exits (cleanup, persist state)
• Notification - During alerts (desktop notifications, logging)
• Stop / SubagentStop - When responses finish (cleanup, summary)
• PreCompact - Before context compaction (save important context)

Common mistake: Using PostToolUse for validation (too late—tool already ran). Use PreToolUse to block operations.

Which Tools Should Trigger This Hook?

Be specific. matcher: "*" runs on every tool call.

Good matchers:

• "Write" - Only file writes
• "Edit|Write" - File modifications
• "Bash" - Shell commands
• "mcp__github__*" - All GitHub MCP tools

Bad matchers:

• "*" - Everything (use only for logging/metrics)

What Input Does This Hook Need?

Different tools provide different input. Check what's available:

# PreToolUse / PostToolUse
{
  "input": {
    "file_path": "/path/to/file.ts",     // Read, Write, Edit
    "command": "npm test",                // Bash
    "old_string": "...",                  // Edit
    "new_string": "..."                   // Edit
  }
}

Validate fields exist before using them:

FILE=$(echo "$INPUT" | jq -r '.input.file_path // empty')
if [[ -z "$FILE" ]]; then
  echo "No file path provided" >&2
  exit 1
fi

Should This Be a Command Hook or Prompt Hook?

Command hooks (type: "command"):

• Fast (milliseconds)
• Deterministic
• Good for: formatting, logging, file checks

Prompt hooks (type: "prompt"):

• Slow (2-10 seconds)
• Context-aware (uses LLM)
• Good for: complex validation, security analysis, intent detection

Rule of thumb: Use command hooks unless you need LLM reasoning.

What Exit Code Communicates Success/Failure?

• exit 0 - Success (continue operation)
• exit 2 - Block operation (show error to Claude)
• exit 1 or other - Non-blocking error (log but continue)

For PreToolUse hooks:

• Exit 2 blocks the tool from running
• Exit 0 allows it (optionally with modified input)

For PostToolUse hooks:

• Exit codes don't block (tool already ran)
• Use exit 0 for success, 1 for logging errors

Top 5 Pitfalls (Must Know)

Pitfall #1: Not Quoting Variables

Error: Hooks break on filenames with spaces or special characters

Why: Unquoted variables split on whitespace

Example:

# ❌ WRONG - breaks on "my file.txt"
cat $FILE
prettier --write $FILE
rm $FILE

# ✅ RIGHT - handles spaces and special chars
cat "$FILE"
prettier --write "$FILE"
rm "$FILE"

Why this matters: Files with spaces ("my file.txt"), Unicode ("文件.txt"), or special chars ("file (1).txt") are common.

For quoting best practices: Load references/security-requirements.md for comprehensive input handling patterns.

Pitfall #2: Trusting Input Without Validation

Error: Hook executes on malicious or malformed input

Why: Not validating JSON fields before using them

Example:

# ❌ DANGEROUS - no validation
FILE=$(jq -r '.input.file_path')
rm "$FILE"  # Could delete ../../../etc/passwd

# ✅ SAFE - validate first
FILE=$(jq -r '.input.file_path // empty')
[[ -n "$FILE" ]] || exit 1
[[ "$FILE" == "$CLAUDE_PROJECT_DIR"* ]] || exit 2
[[ "$FILE" != *".."* ]] || exit 2
rm "$FILE"

Why this matters: Prevents path traversal attacks, protects files outside project, prevents malformed input crashes.

For complete security patterns: Load references/security-requirements.md.

Pitfall #3: Blocking Operations Too Long

Error: Hook takes 30+ seconds, blocking Claude

Why: Running expensive operations (tests, builds) synchronously in hook

Example:

# ❌ BLOCKS Claude for 30 seconds
npm test
npm run build

# ✅ RUN IN BACKGROUND - returns immediately
(npm test > /tmp/test-results.log 2>&1 &)
(npm run build > /tmp/build.log 2>&1 &)
exit 0

Why this matters: Slow hooks create bad user experience. Target < 100ms for PreToolUse, < 500ms for PostToolUse.

For performance optimization: Load references/reliability-performance.md.

Pitfall #4: Wrong Exit Code for Blocking

Error: PreToolUse hook doesn't actually block the operation

Why: Using exit 1 instead of exit 2

Example:

# ❌ WRONG - logs error but doesn't block
if [[ $FILE == ".env" ]]; then
  echo "Don't edit .env" >&2
  exit 1  # Tool still runs!
fi

# ✅ RIGHT - actually blocks
if [[ $FILE == ".env" ]]; then
  echo "Blocked: .env is protected" >&2
  exit 2  # Tool is blocked
fi

Why this matters: Exit 1 only logs errors. Exit 2 is required to block in PreToolUse hooks.

For exit code patterns: Load references/hook-templates.md for complete hook response patterns.

Pitfall #5: Assuming Tools Exist

Error: Hook crashes when dependency is missing

Why: Not checking if tool is installed before using

Example:

# ❌ BREAKS if prettier not installed
prettier --write "$FILE"

# ✅ SAFE - check first
if command -v prettier &>/dev/null; then
  prettier --write "$FILE"
else
  echo "prettier not installed, skipping" >&2
  exit 0  # Success exit, just skip
fi

Why this matters: Users may not have all tools installed. Hooks should degrade gracefully.

For reliability patterns: Load references/reliability-performance.md.

Critical Rules

Always Do

✅ Validate all JSON input before using (jq -r '... // empty') ✅ Quote all variables containing paths or user input ✅ Use absolute paths for scripts (${CLAUDE_PLUGIN_ROOT}/...) ✅ Block sensitive files (.env, *.key, credentials) ✅ Check if required tools exist (command -v toolname) ✅ Set reasonable timeouts (< 5s for PreToolUse) ✅ Run heavy operations in background ✅ Test with edge cases (spaces, Unicode, special chars) ✅ Use exit 2 to block in PreToolUse hooks ✅ Log errors to stderr or file, not stdout

Never Do

❌ Trust JSON input without validation ❌ Use unquoted variables ($FILE instead of "$FILE") ❌ Use relative paths for scripts ❌ Skip path sanitization (check for .., validate in project) ❌ Assume tools are installed ❌ Block for > 1 second in PreToolUse hooks ❌ Use exit 1 when you mean to block (use exit 2) ❌ Log sensitive data to stdout or files ❌ Use matcher: "*" unless truly necessary

When to Load References

Load reference files when working on specific hook aspects:

Security Requirements (references/security-requirements.md)

Load when:

• Implementing input validation and sanitization
• Securing hooks that handle sensitive data
• Blocking sensitive files (.env, keys, credentials)
• Preventing path traversal attacks
• Understanding security vulnerabilities and best practices
• Testing security with malicious input

Reliability & Performance (references/reliability-performance.md)

Load when:

• Handling missing dependencies or tools
• Setting timeouts and handling slow operations
• Optimizing hook performance (< 100ms target)
• Running heavy operations in background
• Caching expensive results
• Testing with edge cases (Unicode, spaces, deep paths)
• Deduplicating expensive operations

Code Templates (references/code-templates.md)

Load when:

• Starting a new hook and need working examples
• Implementing format-on-save functionality
• Blocking sensitive files from modification
• Logging commands or operations
• Using prompt-based security analysis
• Customizing templates for specific use cases

Testing & Debugging (references/testing-debugging.md)

Load when:

• Writing test cases for hooks
• Debugging hook failures or unexpected behavior
• Testing with edge cases (malformed JSON, missing fields)
• Checking hook execution in transcript (Ctrl-R)
• Profiling hook performance
• Creating automated test suites

Publishing Guide (references/publishing-guide.md)

Load when:

• Publishing hooks to PRPM registry
• Creating package manifest (prpm.json)
• Configuring hook.json with advanced options
• Using continue, stopReason, suppressOutput, systemMessage
• Writing README.md for users
• Understanding versioning and publishing commands

Quick Reference (references/quick-reference.md)

Load when:

• Need quick syntax lookup (exit codes, jq patterns)
• Looking up environment variables
• Finding common bash patterns (file validation, background execution)
• Checking hook events and matchers
• Need performance tips summary
• Looking up JSON input structure

Final Checklist

Before publishing a hook:

• Validates all stdin input with jq
• Quotes all variables
• Uses absolute paths for scripts
• Blocks sensitive files (.env, *.key, etc.)
• Handles missing tools gracefully
• Sets reasonable timeout (< 5s for PreToolUse)
• Logs errors to stderr or file, not stdout
• Tests with edge cases (spaces, Unicode, malformed JSON)
• Tests in real Claude Code session
• Documents dependencies in README
• Uses semantic versioning
• Clear description and tags

Using Bundled Resources

This skill includes 6 reference files for on-demand loading:

Security & Reliability (2 files):

• security-requirements.md - Input validation, path sanitization, blocking sensitive files
• reliability-performance.md - Error handling, timeouts, performance optimization

Implementation (2 files):

• code-templates.md - Working hook examples (format-on-save, block-sensitive, logger, etc.)
• quick-reference.md - Fast syntax lookup (exit codes, jq patterns, environment vars)

Testing & Publishing (2 files):

• testing-debugging.md - Test patterns, edge cases, debugging techniques
• publishing-guide.md - PRPM packaging, advanced configuration, README template

Load references on-demand when specific knowledge is needed. See "When to Load References" section for triggers.

Resources

• Claude Code Hooks Docs
• PRPM Hook Packages

Last verified: 2025-12-17 | Version: 2.0.0