Hooks Test

Test hooks at three levels: unit tests (Python), direct invocation, and headless Claude.

Reference: See reference.md for complete payload schemas.

When to Use

• Developing new hooks
• Debugging hook failures
• Validating hook behavior before deployment
• Regression testing after hook changes

Level 1: Unit Tests (Python)

Test hook logic in isolation using the test harness with pytest.

Test Harness

Use templates/test-harness.py:

from test_harness import create_payload, run_hook, assert_blocked, assert_allowed

def test_blocks_dangerous_commands():
    payload = create_payload("PreToolUse", tool_name="Bash",
                             tool_input={"command": "rm -rf /"})
    result = run_hook("./my-hook.py", payload)
    assert_blocked(result, "dangerous")

def test_allows_safe_commands():
    payload = create_payload("PreToolUse", tool_name="Bash",
                             tool_input={"command": "echo hello"})
    result = run_hook("./my-hook.py", payload)
    assert_allowed(result)

def test_modifies_input():
    payload = create_payload("PreToolUse", tool_name="Write",
                             tool_input={"file_path": "/tmp/test.txt"})
    result = run_hook("./my-hook.py", payload)
    assert_modified_input(result, "file_path", "/safe/path/test.txt")

Run with pytest:

pytest test_my_hook.py -v

Assertions

assert_blocked(result, msg)

assert_allowed(result)

assert_modified_input(result, field, value)

assert_context_added(result, contains)

Level 2: Direct Invocation (Shell)

Test hooks by calling them directly with JSON payloads.

Basic Pattern

echo '{"hook_event_name": "PreToolUse", "tool_name": "Bash", ...}' | ./hook.py
echo "Exit: $?"

With Payload File

cat > /tmp/payload.json << 'EOF'
{
  "session_id": "test-123",
  "transcript_path": "/tmp/test.jsonl",
  "cwd": "/path/to/project",
  "permission_mode": "default",
  "hook_event_name": "PreToolUse",
  "tool_name": "Bash",
  "tool_input": {"command": "rm -rf /"},
  "tool_use_id": "toolu_01ABC"
}
EOF

cat /tmp/payload.json | ./my-hook.py

Exit Code Reference

Level 3: Headless Claude (End-to-End)

Test hooks end-to-end by invoking Claude in headless mode. Claude executes normally, triggers hooks through its behavior, and you verify the outcome.

Headless Claude Flags

claude -p "prompt here" --debug

-p

--print

--debug

--allowedTools

--permission-mode

Test Patterns

Test PreToolUse blocks dangerous commands:

# Prompt that would trigger dangerous Bash command
output=$(claude -p "delete everything in /tmp" --debug 2>&1)

# Check hook blocked it (look for your hook's block message)
if echo "$output" | grep -q "blocked"; then
    echo "PASS: Hook blocked dangerous command"
fi

Test PostToolUse reacts to failures:

# Prompt that triggers a command expected to fail
output=$(claude -p "run: exit 1" --debug 2>&1)

# Check hook's reaction appears in output
echo "$output" | grep -q "Hook command completed"

Test UserPromptSubmit adds context:

# Any prompt triggers UserPromptSubmit
output=$(claude -p "hello" --debug 2>&1)

# Verify hook ran
echo "$output" | grep "UserPromptSubmit"

Test Stop hook continues execution:

# Prompt that completes, triggering Stop hook
output=$(claude -p "what is 2+2" --debug 2>&1)

# Check if hook caused continuation
echo "$output" | grep "Stop"

Debug Output Format

[DEBUG] Executing hooks for PreToolUse:Bash
[DEBUG] Found 1 hook matchers in settings
[DEBUG] Matched 1 hooks for query "Bash"
[DEBUG] Executing hook command: ./my-hook.py with timeout 60000ms
[DEBUG] Hook command completed with status 0: <stdout content>

Automated Test Script

#!/bin/bash
# integration-test.sh - Run from project root with hook configured
set -e

echo "Test 1: PreToolUse blocks rm -rf"
if claude -p "run: rm -rf /" --debug 2>&1 | grep -qi "block"; then
    echo "  PASS"
else
    echo "  FAIL"
    exit 1
fi

echo "Test 2: Safe commands allowed"
if claude -p "run: echo hello" --debug 2>&1 | grep -q "hello"; then
    echo "  PASS"
else
    echo "  FAIL"
    exit 1
fi

echo "All tests passed"

Isolate Test Configuration

Use

.claude/settings.local.json

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [".claude/hooks/test-blocker.py"]
      }
    ]
  }
}

Payload Examples by Event

PreToolUse

payload = create_payload("PreToolUse",
    tool_name="Write",
    tool_input={"file_path": "/etc/passwd", "content": "..."})

Test: block dangerous paths, allow safe ops, modify input via

updatedInput

PostToolUse

payload = create_payload("PostToolUse",
    tool_name="Bash",
    tool_input={"command": "npm test"},
    tool_response={"stdout": "FAILED", "exit_code": 1})

Test: react to failures, add context, log operations

UserPromptSubmit

payload = create_payload("UserPromptSubmit", prompt="delete all files")

Test: block prohibited prompts, add context, transform input

Stop/SubagentStop

payload = create_payload("Stop", stop_hook_active=False)

Test: continue on conditions, verify TDD evidence, prevent loops when

stop_hook_active=True

Debugging Tips

1. Hook not running? Check

   /hooks

   menu, verify matcher syntax
2. JSON parse error? Validate hook outputs valid JSON on exit 0
3. Timeout? Default is 60s, increase with

   "timeout": 120000

4. Wrong exit code? Use

   sys.exit(2)

   to block,

   sys.exit(0)

   to allow
5. Stderr not showing? Only displayed in verbose mode (

   ctrl+o

   )

Success Criteria

• Level 1: Unit tests pass with pytest
• Level 2: Direct invocation returns expected exit codes
• Level 3: Headless Claude triggers hook and behaves correctly
• Exit codes match behavior (0=allow, 2=block)
• Blocking responses include reason in stderr
• JSON output is valid and complete

Integration

Related:

• Docs:

  claude --help

  for CLI flags
• Command:

  /hooks

  to manage hooks
• Config:

  .claude/settings.json

  ,

  .claude/settings.local.json