Plugin Self-Testing

Overview

Comprehensive testing framework for validating plugin components. Tests hooks, agents, skills, commands, and routing logic to ensure everything works as expected.

Announce at start: "I'm using the plugin-test skill to run self-tests."

Test Categories

1. Structure Tests

• All referenced files exist
• YAML frontmatter is valid
• Required fields present
• No broken references

2. Hook Tests

• JSON stdin/stdout protocol works
• Hooks respond without blocking
• Error handling is graceful
• Timeout behavior correct

3. Agent Tests

• Agent definitions load correctly
• Tool permissions are valid
• Routing keywords resolve properly
• Confidence scoring works

4. Skill Tests

• SKILL.md format valid
• Description includes "Use when..."
• Referenced skills exist
• No circular dependencies

5. Command Tests

• Commands resolve to skills
• Arguments parsed correctly
• Help text present
• Output format valid

Test Protocol

Running Tests

# All tests
/plugin-test

# Specific category
/plugin-test hooks
/plugin-test agents
/plugin-test skills
/plugin-test routing

Test Output Format

[PASS] hook/pre-tool-use - JSON response valid
[PASS] hook/post-tool-use - JSON response valid
[FAIL] hook/notification - Missing timestamp field
[SKIP] agent/rapid-prototyper - No test file

Summary: 42 passed, 1 failed, 3 skipped

Test File Format

Tests live in tests/ directory:

tests/
  hooks/
    test-pre-tool-use.json     # Input/expected output pairs
    test-post-tool-use.json
  agents/
    test-routing.json          # Routing verification
  skills/
    test-structure.json        # Structure validation

Hook Test Format

{
  "name": "pre-tool-use",
  "tests": [
    {
      "name": "blocks dangerous commands",
      "input": {
        "tool_name": "Bash",
        "tool_input": {"command": "rm -rf /"}
      },
      "expected": {
        "decision": "block"
      }
    },
    {
      "name": "allows safe reads",
      "input": {
        "tool_name": "Read",
        "tool_input": {"file_path": "/tmp/test.txt"}
      },
      "expected": {
        "decision": "allow"
      }
    }
  ]
}

Routing Test Format

{
  "name": "agent-routing",
  "tests": [
    {
      "name": "bug keyword routes to bug-whisperer",
      "prompt": "fix this bug in the authentication code",
      "expected_agent": "bug-whisperer",
      "min_confidence": 0.6
    }
  ]
}

Execution

1. Load test files from tests/ directory
2. Execute each test:
   • Prepare input
   • Run component (hook/agent/skill)
   • Compare output to expected
3. Report results with pass/fail counts
4. Save report to tests/results/latest.json

Key Principles

• Non-destructive - Tests must not modify production state
• Isolated - Each test runs independently
• Deterministic - Same input = same output
• Fast - Complete suite under 30 seconds
• Informative - Clear failure messages

Integration

Called by:

• CI/CD pipeline
• Pre-commit hook (optional)
• Manual validation

Outputs:

• Console test results
• tests/results/latest.json
• tests/results/history.json (append)