Claude Code Plugins

Community-maintained marketplace

Feedback

typescript-hook-writer

@pr-pm/prpm
38
0

Expert guidance for developing Claude Code hooks in TypeScript with shared utilities, esbuild compilation, and Vitest testing - distributes compiled JS while maintaining TypeScript development experience

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name typescript-hook-writer
description Expert guidance for developing Claude Code hooks in TypeScript with shared utilities, esbuild compilation, and Vitest testing - distributes compiled JS while maintaining TypeScript development experience

TypeScript Hook Writer

Use this skill when developing Claude Code hooks in TypeScript. This skill ensures you maintain type safety, shared utilities, proper build pipeline, and comprehensive testing while distributing single-file JavaScript bundles to users.

When to Use This Skill

  • Creating new TypeScript hooks for Claude Code
  • Setting up the hooks development environment
  • Adding shared utilities for hooks
  • Writing tests for hooks with Vitest
  • Building hooks for distribution
  • Publishing TypeScript-based hooks as PRPM packages

Why TypeScript for Hooks?

Advantages over bash:

  • Type safety catches errors at compile time
  • Shared utility functions reduce code duplication
  • Better IDE support with autocomplete and refactoring
  • Easier to test with Vitest
  • More readable for complex logic
  • Strong validation with TypeScript interfaces

Trade-offs:

  • Requires build step (esbuild)
  • Slightly larger bundle size (~2-3KB vs bash)
  • Users still just need Node.js (no TypeScript dependency)

When to use TypeScript hooks:

  • Complex input validation or pattern matching
  • Hooks that share common logic
  • Hooks requiring automated testing
  • Teams familiar with TypeScript

When to stick with bash:

  • Simple one-off hooks (< 20 lines)
  • Hooks that just call other CLI tools
  • Extreme performance requirements (though difference is negligible)

Project Structure

packages/hooks/
├── package.json              # Build tooling (esbuild, tsx, vitest)
├── tsconfig.json             # TypeScript configuration
├── vitest.config.ts          # Test configuration
├── scripts/
│   └── build-all-hooks.ts    # Build script (compiles all hooks)
├── shared/
│   ├── types.ts              # Shared TypeScript interfaces
│   ├── hook-utils.ts         # Shared utility functions
│   └── hook-utils.test.ts    # Tests for shared utilities

# Each hook lives in .claude/hooks/
.claude/hooks/
├── my-hook/
│   ├── hook.json             # Hook configuration
│   ├── README.md             # Documentation
│   ├── src/
│   │   ├── hook.ts           # TypeScript source (development)
│   │   ├── hook.test.ts      # Vitest tests
│   │   ├── hook-utils.ts     # Copied shared utilities
│   │   └── types.ts          # Copied shared types
│   └── dist/
│       └── hook.js           # Compiled bundle (distributed)

Setup: packages/hooks Infrastructure

1. package.json

{
  "name": "@prpm/hooks",
  "version": "1.0.0",
  "description": "Build system and shared utilities for PRPM Claude Code hooks",
  "private": true,
  "type": "module",
  "scripts": {
    "build": "tsx scripts/build-all-hooks.ts",
    "build:watch": "tsx scripts/build-all-hooks.ts --watch",
    "test": "vitest",
    "test:watch": "vitest --watch",
    "test:run": "vitest run",
    "test:coverage": "vitest run --coverage"
  },
  "devDependencies": {
    "@types/node": "^20.10.0",
    "esbuild": "^0.19.8",
    "tsx": "^4.7.0",
    "typescript": "^5.3.3",
    "vitest": "^1.0.4"
  }
}

2. tsconfig.json

{
  "compilerOptions": {
    "target": "ES2022",
    "module": "ESNext",
    "lib": ["ES2022"],
    "moduleResolution": "node",
    "esModuleInterop": true,
    "strict": true,
    "skipLibCheck": true,
    "resolveJsonModule": true,
    "types": ["node"]
  },
  "include": ["**/*.ts"],
  "exclude": ["node_modules", "dist"]
}

3. vitest.config.ts

import { defineConfig } from 'vitest/config';

export default defineConfig({
  test: {
    globals: true,
    environment: 'node',
    include: ['**/*.test.ts'],
    coverage: {
      provider: 'v8',
      reporter: ['text', 'json', 'html'],
      exclude: [
        'node_modules/',
        'dist/',
        '**/*.test.ts',
        '**/scripts/**',
      ],
    },
  },
});

4. Build Script: scripts/build-all-hooks.ts

#!/usr/bin/env tsx
/**
 * Build script for compiling all Claude Code hooks to standalone JavaScript bundles
 */

import { build, BuildOptions } from 'esbuild';
import { readdirSync, statSync, existsSync, mkdirSync } from 'fs';
import { join, dirname } from 'path';
import { fileURLToPath } from 'url';

const __filename = fileURLToPath(import.meta.url);
const __dirname = dirname(__filename);

// Root directory is app/ which is 3 levels up from packages/hooks/scripts/
const ROOT_DIR = join(__dirname, '../../..');
const HOOKS_DIR = join(ROOT_DIR, '.claude/hooks');

interface HookInfo {
  name: string;
  srcPath: string;
  distPath: string;
}

/**
 * Find all hooks with TypeScript source files
 */
function findHooks(): HookInfo[] {
  const hooks: HookInfo[] = [];

  if (!existsSync(HOOKS_DIR)) {
    console.error(`Hooks directory not found: ${HOOKS_DIR}`);
    return hooks;
  }

  const entries = readdirSync(HOOKS_DIR);

  for (const entry of entries) {
    const hookPath = join(HOOKS_DIR, entry);

    // Skip non-directories and special directories
    if (!statSync(hookPath).isDirectory() || entry === 'shared') {
      continue;
    }

    const srcPath = join(hookPath, 'src/hook.ts');
    const distPath = join(hookPath, 'dist/hook.js');

    if (existsSync(srcPath)) {
      hooks.push({
        name: entry,
        srcPath,
        distPath,
      });
    }
  }

  return hooks;
}

/**
 * Build a single hook
 */
async function buildHook(hook: HookInfo): Promise<void> {
  const buildOptions: BuildOptions = {
    entryPoints: [hook.srcPath],
    bundle: true,
    platform: 'node',
    target: 'node18',
    outfile: hook.distPath,
    format: 'cjs',
    minify: false, // Keep readable for debugging
    sourcemap: false,
    logLevel: 'error',
    banner: {
      js: '#!/usr/bin/env node',
    },
  };

  try {
    // Ensure dist directory exists
    const distDir = dirname(hook.distPath);
    if (!existsSync(distDir)) {
      mkdirSync(distDir, { recursive: true });
    }

    await build(buildOptions);
    console.log(`✓ Built ${hook.name}`);

    // Make the output file executable
    const { chmodSync } = await import('fs');
    chmodSync(hook.distPath, 0o755);
  } catch (error) {
    console.error(`✗ Failed to build ${hook.name}:`, error);
    throw error;
  }
}

/**
 * Build all hooks
 */
async function buildAll(watch: boolean = false): Promise<void> {
  console.log('🔨 Building PRPM Claude Code hooks...\n');

  const hooks = findHooks();

  if (hooks.length === 0) {
    console.log('No hooks found to build.');
    return;
  }

  console.log(`Found ${hooks.length} hooks:\n`);

  // Build all hooks in parallel
  try {
    await Promise.all(hooks.map(hook => buildHook(hook)));
    console.log(`\n✓ Built ${hooks.length} hooks successfully`);
  } catch (error) {
    console.error('\n✗ Build failed');
    process.exit(1);
  }

  if (watch) {
    console.log('\n👀 Watching for changes...');
    console.log('⚠️  Watch mode not yet implemented. Run `npm run build` after changes.');
  }
}

// Parse command line args
const args = process.argv.slice(2);
const watch = args.includes('--watch') || args.includes('-w');

// Run build
buildAll(watch).catch(error => {
  console.error('Build error:', error);
  process.exit(1);
});

Shared Utilities Pattern

Why Copy Instead of Import?

Each hook gets its own copy of shared utilities rather than importing from a shared package:

Benefits:

  • Each hook is a standalone single-file bundle
  • No external dependencies at runtime
  • Simpler distribution (just one .js file)
  • No module resolution issues
  • Each hook can be updated independently

Trade-off:

  • Slight code duplication (~1-2KB per hook)
  • Changes to shared utilities require rebuilding all hooks

shared/types.ts

Define TypeScript interfaces for hook input, exit codes, and options:

/**
 * Shared TypeScript types for Claude Code hooks
 */

export interface HookInput {
  session_id?: string;
  transcript_path?: string;
  current_dir?: string;
  input?: {
    file_path?: string;
    command?: string;
    content?: string;
    old_string?: string;
    new_string?: string;
    [key: string]: any;
  };
  message?: string;
  tool?: string;
  [key: string]: any;
}

export enum HookExitCode {
  Success = 0,    // Continue operation
  Error = 1,      // Log error but continue
  Block = 2,      // Block operation (PreToolUse only)
}

export interface ExecOptions {
  skipOnMissing?: boolean;  // Exit successfully if command not found
  background?: boolean;     // Run in background (don't wait)
  timeout?: number;         // Timeout in milliseconds
  env?: Record<string, string>; // Environment variables
}

export interface PatternMatch {
  matched: boolean;
  pattern?: string;
}

shared/hook-utils.ts

Common utility functions used across hooks:

import { readFileSync, appendFileSync, existsSync } from 'fs';
import { execSync } from 'child_process';
import type { HookInput, HookExitCode, ExecOptions, PatternMatch } from './types';

/**
 * Read and parse JSON from stdin
 */
export function readStdin(): HookInput {
  try {
    const input = readFileSync(0, 'utf-8');
    return JSON.parse(input);
  } catch (error) {
    return {};
  }
}

/**
 * Extract file path from hook input
 */
export function getFilePath(input: HookInput): string | undefined {
  return input.input?.file_path;
}

/**
 * Extract command from hook input
 */
export function getCommand(input: HookInput): string | undefined {
  return input.input?.command;
}

/**
 * Extract content from hook input
 */
export function getContent(input: HookInput): string | undefined {
  return input.input?.content || input.input?.new_string;
}

/**
 * Check if file has one of the specified extensions
 */
export function hasExtension(filePath: string, extensions: string[]): boolean {
  return extensions.some(ext => filePath.endsWith(ext));
}

/**
 * Check if a command exists in PATH
 */
export function commandExists(command: string): boolean {
  try {
    execSync(`command -v ${command}`, { stdio: 'ignore' });
    return true;
  } catch {
    return false;
  }
}

/**
 * Execute a command with options
 */
export function execCommand(
  command: string,
  args: string[],
  options: ExecOptions = {}
): void {
  // Check if command exists if skipOnMissing is true
  if (options.skipOnMissing && !commandExists(command)) {
    return;
  }

  const fullCommand = `${command} ${args.map(arg => `"${arg}"`).join(' ')}`;

  if (options.background) {
    // Run in background - don't wait for completion
    execSync(`(${fullCommand} &)`, {
      stdio: 'ignore',
      timeout: options.timeout,
      env: { ...process.env, ...options.env },
    });
  } else {
    // Run synchronously
    execSync(fullCommand, {
      stdio: 'inherit',
      timeout: options.timeout,
      env: { ...process.env, ...options.env },
    });
  }
}

/**
 * Match file path against glob patterns
 */
export function matchesPattern(filePath: string, patterns: string[]): PatternMatch {
  for (const pattern of patterns) {
    // Convert glob pattern to regex
    const regexPattern = pattern
      .replace(/\./g, '\\.')
      .replace(/\*/g, '.*')
      .replace(/\?/g, '.');

    if (new RegExp(`^${regexPattern}$`).test(filePath)) {
      return { matched: true, pattern };
    }
  }

  return { matched: false };
}

/**
 * Append line to log file
 */
export function appendToLog(logFile: string, line: string): void {
  try {
    appendFileSync(logFile, line + '\n', 'utf-8');
  } catch {
    // Fail silently
  }
}

/**
 * Get current timestamp in YYYY-MM-DD HH:MM:SS format
 */
export function getTimestamp(): string {
  return new Date().toISOString().replace('T', ' ').substring(0, 19);
}

/**
 * Log error message to stderr
 */
export function logError(message: string): void {
  console.error(message);
}

/**
 * Log warning message to stderr
 */
export function logWarning(message: string): void {
  console.error(message);
}

/**
 * Exit hook with specified code
 */
export function exitHook(code: HookExitCode): never {
  process.exit(code);
}

// Re-export HookExitCode for convenience
export { HookExitCode } from './types';

Creating a TypeScript Hook

Step 1: Create Hook Directory Structure

mkdir -p .claude/hooks/my-hook/src
mkdir -p .claude/hooks/my-hook/dist

Step 2: Copy Shared Utilities

Copy shared/types.ts and shared/hook-utils.ts into the hook's src/ directory:

cp packages/hooks/shared/types.ts .claude/hooks/my-hook/src/
cp packages/hooks/shared/hook-utils.ts .claude/hooks/my-hook/src/

Why copy instead of symlink? Each hook becomes a standalone bundle when compiled. The build process bundles utilities into the final .js file.

Step 3: Write Hook Implementation

.claude/hooks/my-hook/src/hook.ts:

#!/usr/bin/env tsx
/**
 * My Hook
 * Description of what this hook does
 */

import {
  readStdin,
  getFilePath,
  hasExtension,
  execCommand,
  logError,
  logWarning,
  exitHook,
  HookExitCode,
} from './hook-utils';

async function main() {
  // Read input from stdin
  const input = readStdin();

  // Extract file path
  const filePath = getFilePath(input);
  if (!filePath) {
    exitHook(HookExitCode.Success);
  }

  // Validate file extension
  const supportedExtensions = ['.ts', '.tsx', '.js', '.jsx'];
  if (!hasExtension(filePath, supportedExtensions)) {
    exitHook(HookExitCode.Success);
  }

  // Perform hook action
  try {
    execCommand('prettier', ['--write', filePath], {
      skipOnMissing: true,
      background: true,
    });

    exitHook(HookExitCode.Success);
  } catch (error) {
    logError(`Failed to format ${filePath}: ${error}`);
    exitHook(HookExitCode.Error);
  }
}

main().catch(() => {
  exitHook(HookExitCode.Success); // Don't block on errors
});

Step 4: Create hook.json Configuration

.claude/hooks/my-hook/hook.json:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [
          {
            "type": "command",
            "command": "node .claude/hooks/my-hook/dist/hook.js",
            "timeout": 5000
          }
        ]
      }
    ]
  }
}

Important: Reference dist/hook.js (compiled), not src/hook.ts (source).

Step 4.5: Advanced Hook Configuration (Optional)

All hook types support optional fields for controlling execution behavior:

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Edit|Write",
      "hooks": [{
        "type": "command",
        "command": "node .claude/hooks/my-hook/dist/hook.js",
        "timeout": 5000,
        "continue": true,              // Whether Claude continues after hook (default: true)
        "stopReason": "string",        // Message shown when continue is false
        "suppressOutput": false,       // Hide stdout from transcript (default: false)
        "systemMessage": "string"      // Warning message shown to user
      }]
    }]
  }
}

continue (boolean, default: true)

Controls whether Claude continues after hook execution.

When to use false:

  • Security hooks that must block operations
  • Validation hooks that found critical errors
  • Hooks that require user intervention
{
  "type": "command",
  "command": "node .claude/hooks/security-validator/dist/hook.js",
  "continue": false,
  "stopReason": "Security validation failed. Please review the detected issues before proceeding."
}

Exit code interaction:

  • If hook exits with HookExitCode.Block (2): continue is ignored, operation is blocked
  • If hook exits with HookExitCode.Success (0) or HookExitCode.Error (1): continue field determines behavior

stopReason (string)

Message displayed to user when continue: false. Should explain why execution stopped and what action is needed.

{
  "continue": false,
  "stopReason": "Pre-commit checks failed. Fix linting errors and try again."
}

suppressOutput (boolean, default: false)

Hides hook stdout from transcript mode (Ctrl-R). Stderr is always shown.

When to use true:

  • Hooks that produce verbose output
  • Debugging logs not useful to users
  • Noisy background operations
{
  "type": "command",
  "command": "node .claude/hooks/cloud-sync/dist/hook.js",
  "suppressOutput": true  // Don't show sync progress in transcript
}

Note: Always show critical errors via stderr (use logError()), as stderr is never suppressed.

systemMessage (string)

Warning or info message shown to user when hook executes. Useful for non-blocking warnings.

TypeScript example:

// In your hook.ts
if (outdatedDeps.length > 0) {
  logWarning(`Found ${outdatedDeps.length} outdated dependencies`);
  // systemMessage in hook.json will also show to user
}
{
  "type": "command",
  "command": "node .claude/hooks/dependency-checker/dist/hook.js",
  "systemMessage": "⚠️  Some dependencies are outdated. Consider running 'npm update'."
}

Difference from stopReason:

  • systemMessage: Informational, Claude continues
  • stopReason: Critical, requires continue: false

Step 5: Build the Hook

cd packages/hooks
pnpm build

Output:

🔨 Building PRPM Claude Code hooks...

Found 1 hooks:

✓ Built my-hook

✓ Built 1 hooks successfully

The compiled hook is now at .claude/hooks/my-hook/dist/hook.js (~2-3KB single file).

Testing TypeScript Hooks

Step 1: Create Test File

.claude/hooks/my-hook/src/hook.test.ts:

import { describe, it, expect, vi, beforeEach } from 'vitest';

// Mock the hook-utils module
vi.mock('./hook-utils', () => ({
  readStdin: vi.fn(),
  getFilePath: vi.fn(),
  hasExtension: vi.fn(),
  execCommand: vi.fn(),
  logError: vi.fn(),
  exitHook: vi.fn((code: number) => {
    throw new Error(`EXIT_${code}`);
  }),
  HookExitCode: {
    Success: 0,
    Error: 1,
    Block: 2,
  },
}));

describe('my-hook', () => {
  beforeEach(() => {
    vi.clearAllMocks();
  });

  it('should exit successfully if no file path is provided', async () => {
    const { readStdin, getFilePath, exitHook, HookExitCode } = await import('./hook-utils');

    vi.mocked(readStdin).mockReturnValue({});
    vi.mocked(getFilePath).mockReturnValue(undefined);

    try {
      await import('./hook');
    } catch (error: any) {
      expect(error.message).toBe('EXIT_0');
    }

    expect(exitHook).toHaveBeenCalledWith(HookExitCode.Success);
  });

  it('should format supported file types', async () => {
    const { readStdin, getFilePath, hasExtension, execCommand, exitHook, HookExitCode } = await import('./hook-utils');

    vi.mocked(readStdin).mockReturnValue({
      input: { file_path: '/path/to/file.ts' },
    });
    vi.mocked(getFilePath).mockReturnValue('/path/to/file.ts');
    vi.mocked(hasExtension).mockReturnValue(true);

    try {
      await import('./hook');
    } catch (error: any) {
      expect(error.message).toBe('EXIT_0');
    }

    expect(execCommand).toHaveBeenCalledWith(
      'prettier',
      ['--write', '/path/to/file.ts'],
      expect.objectContaining({ background: true })
    );
  });

  it('should skip unsupported file types', async () => {
    const { readStdin, getFilePath, hasExtension, execCommand, exitHook } = await import('./hook-utils');

    vi.mocked(readStdin).mockReturnValue({
      input: { file_path: '/path/to/file.py' },
    });
    vi.mocked(getFilePath).mockReturnValue('/path/to/file.py');
    vi.mocked(hasExtension).mockReturnValue(false);

    try {
      await import('./hook');
    } catch (error: any) {
      // Expected exit
    }

    expect(execCommand).not.toHaveBeenCalled();
  });
});

Step 2: Run Tests

cd packages/hooks
pnpm test:run

Output:

✓ .claude/hooks/my-hook/src/hook.test.ts (3 tests) 5ms

Test Files  1 passed (1)
Tests  3 passed (3)
Duration  182ms

Step 3: Run Tests with Coverage

pnpm test:coverage

Coverage report shows which code paths are tested.

Build Workflow: TypeScript → JavaScript

Development vs Distribution

Development files (NOT distributed):

  • .claude/hooks/my-hook/src/hook.ts - TypeScript source
  • .claude/hooks/my-hook/src/hook.test.ts - Tests
  • .claude/hooks/my-hook/src/types.ts - Type definitions
  • .claude/hooks/my-hook/src/hook-utils.ts - Utilities

Distribution files (what users get):

  • .claude/hooks/my-hook/dist/hook.js - Compiled JavaScript bundle (~2-3KB)
  • .claude/hooks/my-hook/hook.json - Hook configuration
  • .claude/hooks/my-hook/README.md - Documentation

Build Process

The build script (packages/hooks/scripts/build-all-hooks.ts) does the following:

  1. Scans .claude/hooks/ for directories with src/hook.ts
  2. Compiles each hook with esbuild:
    • Bundles all imports into single file
    • Converts TypeScript to JavaScript
    • Targets Node.js 18+
    • Outputs CommonJS format
    • Adds #!/usr/bin/env node shebang
  3. Outputs to dist/hook.js in each hook directory
  4. Sets permissions to make file executable (chmod +x)

When to Build

Automatic build:

  • Publishing with prpm publish - prepublishOnly script builds automatically (if configured)

Manual build for:

  • Testing hook locally before committing
  • Debugging compiled output
  • Verifying build succeeds before pushing

How to build manually:

# Build all hooks once
cd packages/hooks
pnpm build

# Output:
# 🔨 Building PRPM Claude Code hooks...
# Found 7 hooks:
# ✓ Built prettier-on-save
# ✓ Built command-logger
# ...
# ✓ Built 7 hooks successfully

Build output structure:

.claude/hooks/my-hook/
├── src/
│   ├── hook.ts           ← TypeScript source (input)
│   ├── hook-utils.ts
│   └── types.ts
└── dist/
    └── hook.js           ← Compiled JavaScript (output)

What gets bundled:

esbuild traces all imports and bundles them into a single dist/hook.js:

// src/hook.ts imports these
import { readStdin, getFilePath } from './hook-utils';
import { HookExitCode } from './types';

// dist/hook.js contains:
// - All code from hook.ts
// - All code from hook-utils.ts
// - All type definitions (compiled to runtime checks)
// - No external dependencies
// - Total size: ~2-3KB

Why this approach:

  • ✅ Users don't need TypeScript or build tools
  • ✅ Single-file distribution is simple
  • ✅ No runtime dependencies (just Node.js)
  • ✅ Hooks load instantly (no module resolution)
  • ✅ Each hook is independent

Publishing TypeScript Hooks as PRPM Packages

Step 1: Build Hooks

IMPORTANT: Always build before updating prpm.json or publishing:

cd packages/hooks
pnpm build

Verify dist files exist:

ls -lh .claude/hooks/*/dist/hook.js
# Should show compiled hooks with ~2-3KB size each

Step 2: Update prpm.json

Add the hook to the root prpm.json:

{
  "packages": [
    {
      "name": "my-hook",
      "version": "1.0.0",
      "description": "Brief description of what the hook does",
      "format": "claude",
      "subtype": "hook",
      "tags": ["formatting", "automation", "typescript"],
      "files": [
        ".claude/hooks/my-hook/hook.json",
        ".claude/hooks/my-hook/dist/hook.js",
        ".claude/hooks/my-hook/README.md"
      ]
    }
  ]
}

Important files array:

  • hook.json - Hook configuration
  • dist/hook.js - Compiled JavaScript (NOT src/hook.ts)
  • README.md - Documentation

Do NOT include:

  • src/ directory (source code)
  • *.test.ts files
  • node_modules/
  • Development files

Step 2: Create README.md

.claude/hooks/my-hook/README.md:

# My Hook

Brief description of what this hook does.

## What It Does

- Automatically formats code after Claude edits files
- Supports TypeScript, JavaScript, JSON, and Markdown
- Runs in background (non-blocking)
- Gracefully skips if Prettier not installed

## Installation

```bash
prpm install @prpm/my-hook

Requirements

  • Node.js 18+ (already required for Claude Code)
  • Prettier (optional): npm install -g prettier

Configuration

This hook activates on PostToolUse for Edit and Write tools.

To customize supported file extensions, fork and modify the source.

Examples

When Claude writes a TypeScript file:

Claude: I'll create a new component...
[Hook auto-formats component.tsx with Prettier]

Troubleshooting

Hook not running?

  • Check .claude/settings.json includes the hook
  • Verify dist/hook.js exists and is executable
  • Check transcript (Ctrl-R) for hook errors

Format not applying?

  • Ensure Prettier is installed: prettier --version
  • Check Prettier config in project (.prettierrc)

### Step 3: Set Up Automatic Build Before Publishing

**Good news:** PRPM now supports `prepublishOnly` scripts! Add this to your prpm.json:

```json
{
  "name": "prpm-packages",
  "license": "MIT",
  "repository": "https://github.com/username/repo",
  "scripts": {
    "prepublishOnly": "cd packages/hooks && npm run build"
  },
  "packages": [
    // ... your packages
  ]
}

What happens:

  • When you run prpm publish, the prepublishOnly script runs automatically
  • Hooks are compiled from TypeScript to JavaScript
  • If the build fails, publish is aborted (prevents publishing broken code)
  • If the build succeeds, publishing continues with up-to-date dist files

Manual build (for local testing):

cd packages/hooks
npm run build

Verify dist/hook.js files exist:

ls -lh .claude/hooks/*/dist/hook.js
# Should show compiled hooks with ~2-3KB size each

Step 4: Publish

# From project root
prpm publish

What happens automatically:

  1. prepublishOnly script runs: cd packages/hooks && npm run build
  2. All hooks are compiled to dist/hook.js
  3. Packages are published with up-to-date compiled files

Users will receive:

  • hook.json - Configuration
  • dist/hook.js - Single-file executable (~2-3KB)
  • README.md - Documentation

Users do NOT get:

  • TypeScript source (src/ directory)
  • Build tooling
  • Tests

Why this matters: The prepublishOnly script prevents publishing stale dist files. If you modify a hook's TypeScript source but forget to build, the build happens automatically before publish.

Best Practices

1. Keep Hooks Fast

Target < 100ms execution time. Use background execution for slow operations:

// BAD - blocks for 5 seconds
execCommand('npm', ['test']);

// GOOD - runs in background
execCommand('npm', ['test'], { background: true });

2. Fail Gracefully

Never crash. Handle missing tools:

execCommand('prettier', ['--write', filePath], {
  skipOnMissing: true,  // Exit successfully if prettier not found
  background: true,
});

3. Use Type Guards

Validate input shape:

function isValidInput(input: HookInput): boolean {
  return !!(input.input?.file_path && typeof input.input.file_path === 'string');
}

if (!isValidInput(input)) {
  exitHook(HookExitCode.Success);
}

4. Copy Shared Utilities

Always copy (not import) shared utilities into each hook's src/ directory:

cp packages/hooks/shared/{types.ts,hook-utils.ts} .claude/hooks/my-hook/src/

This ensures standalone compilation.

5. Test Edge Cases

Test with edge cases:

it('should handle files with spaces', async () => {
  const input = { input: { file_path: '/path/my file.ts' } };
  // ...
});

it('should handle Unicode filenames', async () => {
  const input = { input: { file_path: '/path/文件.ts' } };
  // ...
});

it('should handle missing input fields', async () => {
  const input = { input: {} };
  // ...
});

6. Use Descriptive Exit Codes

// Success - continue operation
exitHook(HookExitCode.Success);

// Block - prevent operation (PreToolUse only)
exitHook(HookExitCode.Block);

// Error - log but continue
exitHook(HookExitCode.Error);

7. Log to stderr

// WRONG - pollutes transcript
console.log('Processing file...');

// RIGHT - logs to stderr
logError('⚠️  Warning: something happened');
logWarning('ℹ️  Info: skipping file');

Common Patterns

Pattern: File Extension Filter

const supportedExtensions = ['.ts', '.tsx', '.js', '.jsx'];

if (!hasExtension(filePath, supportedExtensions)) {
  exitHook(HookExitCode.Success);
}

Pattern: Sensitive File Blocker

const blockedPatterns = ['.env', '.env.*', '*.pem', '*.key', '*credentials*'];

const match = matchesPattern(filePath, blockedPatterns);
if (match.matched) {
  logError(`⛔ Blocked: Cannot modify sensitive file '${filePath}'`);
  logError(`   Pattern: ${match.pattern}`);
  exitHook(HookExitCode.Block);
}

Pattern: Command Logger

import { join } from 'path';
import { homedir } from 'os';

const command = getCommand(input);
if (!command) {
  exitHook(HookExitCode.Success);
}

const logFile = join(homedir(), '.claude-commands.log');
const logLine = `[${getTimestamp()}] ${command}`;

appendToLog(logFile, logLine);
exitHook(HookExitCode.Success);

Pattern: Content Validator

const content = getContent(input);
if (!content) {
  exitHook(HookExitCode.Success);
}

const dangerousPatterns = [
  /password\s*=\s*["'][^"']+["']/i,
  /api[_-]?key\s*=\s*["'][^"']+["']/i,
  /AKIA[0-9A-Z]{16}/, // AWS access key
];

for (const pattern of dangerousPatterns) {
  if (pattern.test(content)) {
    logWarning(`⚠️  Warning: Potential credential detected in ${filePath}`);
    logWarning(`   Pattern matched: ${pattern}`);
    break;
  }
}

exitHook(HookExitCode.Success);

Debugging TypeScript Hooks

1. Test Compilation

cd packages/hooks
pnpm build

Check for TypeScript errors.

2. Test Execution Manually

echo '{"input":{"file_path":"/tmp/test.ts"}}' | node .claude/hooks/my-hook/dist/hook.js
echo $?  # Check exit code

3. Check Hook Registration

Verify hook appears in .claude/settings.json:

cat .claude/settings.json | jq '.hooks'

4. View Transcript

Run Claude Code and check transcript (Ctrl-R) for hook execution:

PostToolUse hook: my-hook
  command: node .claude/hooks/my-hook/dist/hook.js
  exit: 0
  duration: 47ms

5. Add Debug Logging

Temporarily add debug output:

logError(`[DEBUG] Processing file: ${filePath}`);
logError(`[DEBUG] Extensions: ${JSON.stringify(supportedExtensions)}`);

Migration: Bash to TypeScript

Converting an existing bash hook to TypeScript:

Before (bash)

#!/bin/bash
set -euo pipefail

INPUT=$(cat)
FILE=$(echo "$INPUT" | jq -r '.input.file_path // empty')

if [[ -z "$FILE" ]]; then
  exit 0
fi

if ! command -v prettier &>/dev/null; then
  exit 0
fi

prettier --write "$FILE" &
exit 0

After (TypeScript)

#!/usr/bin/env tsx
import {
  readStdin,
  getFilePath,
  execCommand,
  exitHook,
  HookExitCode,
} from './hook-utils';

async function main() {
  const input = readStdin();
  const filePath = getFilePath(input);

  if (!filePath) {
    exitHook(HookExitCode.Success);
  }

  execCommand('prettier', ['--write', filePath], {
    skipOnMissing: true,
    background: true,
  });

  exitHook(HookExitCode.Success);
}

main().catch(() => exitHook(HookExitCode.Success));

Benefits:

  • Type safety for input object
  • Shared utilities reduce code
  • Easier to test
  • Better IDE support

Quick Reference

Build Commands

pnpm build              # Build all hooks once
pnpm build:watch        # Watch mode (not yet implemented)
pnpm test               # Run tests in watch mode
pnpm test:run           # Run tests once
pnpm test:coverage      # Run tests with coverage

Hook Configuration Fields

Required:

  • type - "command" or "prompt"
  • command - Path to compiled hook (e.g., "node .claude/hooks/my-hook/dist/hook.js")

Optional:

  • timeout - Max execution time in ms (default: 60000)
  • continue - Continue after hook? (default: true)
  • stopReason - Message when continue=false
  • suppressOutput - Hide stdout from transcript (default: false)
  • systemMessage - Warning message to user

Exit Codes (HookExitCode enum)

HookExitCode.Success = 0    // Continue operation
HookExitCode.Error = 1      // Log error but continue
HookExitCode.Block = 2      // Block operation (PreToolUse only)

Hook Structure Checklist

  • Created .claude/hooks/my-hook/src/hook.ts
  • Copied types.ts and hook-utils.ts to src/
  • Created hook.json referencing dist/hook.js
  • Created README.md with installation and usage
  • Created hook.test.ts with test coverage
  • Built hook with pnpm build
  • Verified dist/hook.js exists and is executable
  • Added to prpm.json with correct files array
  • Tested manually with sample JSON input
  • Tested in real Claude Code session

prpm.json Files Array

Include:

  • .claude/hooks/my-hook/hook.json
  • .claude/hooks/my-hook/dist/hook.js
  • .claude/hooks/my-hook/README.md

Exclude:

  • src/ directory
  • *.test.ts files
  • node_modules/
  • Development files

Common Utilities

readStdin()                          // Parse stdin JSON
getFilePath(input)                   // Extract file path
getCommand(input)                    // Extract command
getContent(input)                    // Extract content
hasExtension(path, ['.ts', '.js'])   // Check extension
matchesPattern(path, ['*.env'])      // Glob matching
commandExists('prettier')            // Check command exists
execCommand('cmd', ['arg'], opts)    // Execute command
appendToLog(file, line)              // Append to log
getTimestamp()                       // Current timestamp
logError(msg)                        // Log to stderr
logWarning(msg)                      // Log warning
exitHook(HookExitCode.Success)       // Exit with code

Resources