Test Agents and Skills Command

You are running tests to verify that agents and skills are invoked correctly when prompts match their "when to use" criteria.

Quick Start

Run ALL tests with a single command:

python .claude/tests/run_all_tests.py

Test Suites

The comprehensive test runner includes:

1. Unit Tests - Tests for init-project.py functionality
2. Init-Project Tests - Verifies project.json and symlinks
3. Visibility Tests - Verifies Claude can see agents/commands/skills
4. CLAUDE.md Tests - Verifies documentation is generated correctly
5. Agent Trigger Tests - Tests each ACTIVE agent is triggered by matching prompts
6. Skill Trigger Tests - Tests each ACTIVE skill is triggered by matching prompts

Command Options

# Run ALL tests
python .claude/tests/run_all_tests.py

# Run specific test suites
python .claude/tests/run_all_tests.py --unit          # Unit tests only
python .claude/tests/run_all_tests.py --init          # Init-project tests only
python .claude/tests/run_all_tests.py --visibility    # Visibility tests only
python .claude/tests/run_all_tests.py --claude-md     # CLAUDE.md tests only
python .claude/tests/run_all_tests.py --triggers      # Agent + skill trigger tests
python .claude/tests/run_all_tests.py --agents        # Agent trigger tests only
python .claude/tests/run_all_tests.py --skills        # Skill trigger tests only

# Verbose output
python .claude/tests/run_all_tests.py --verbose

Test Definitions

Test definitions are stored in JSON files:

• .claude/tests/agents/test_definitions.json - 16 agent test definitions (ALL agents)
• .claude/tests/skills/test_definitions.json - 54 skill test definitions (ALL skills)

Important: ALL agents/skills have test definitions, but only ACTIVE ones (symlinked) are tested when running.

How Trigger Tests Work

For each active agent/skill:

1. Send a test prompt to Claude CLI (claude -p "prompt" --dangerously-skip-permissions)
2. Check .claude/logs/agent-invocations.log or .claude/logs/skills.log
3. Verify the correct agent/skill was invoked
4. Report PASS/FAIL

Expected Output

======================================================================
 COMPREHENSIVE TEST SUITE FOR /init-project
======================================================================

  Project root: /path/to/project
  Logs dir: /path/to/project/.claude/logs
  Active agents: 9
  Active skills: 21
  Active commands: 12

======================================================================
 UNIT TESTS
======================================================================
  Passed: 34
  Failed: 0
  Skipped: 8

======================================================================
 INIT-PROJECT TESTS
======================================================================
  [✓ PASS] project.json exists - project.json already exists
  [✓ PASS] Symlinks created - Symlinks exist in agents/, skills/, commands/
  [✓ PASS] Symlinks match categories - Symlinks match categories: dev, claude, docs

...

======================================================================
 FINAL TEST SUMMARY
======================================================================
  ✓ Unit Tests: 34 passed, 0 failed, 8 skipped
  ✓ Init-Project Tests: 3 passed, 0 failed
  ✓ Visibility Tests: 3 passed, 0 failed
  ✓ CLAUDE.md Tests: 2 passed, 0 failed
  ✓ Agent Trigger Tests: 9 passed, 0 failed, 7 skipped
  ✓ Skill Trigger Tests: 21 passed, 0 failed, 33 skipped

  TOTAL: 72 passed, 0 failed, 48 skipped

  ✓ ALL TESTS PASSED!

Instructions

1. Run python .claude/tests/run_all_tests.py to execute all tests
2. Review the output for any failures
3. If trigger tests fail, check that:
   • The agent/skill is properly symlinked
   • The logging hooks are configured in .claude/settings.json
   • The test prompt matches the agent/skill's "when to use" criteria