Reactive Watchers
Monitor agent execution in real-time with fail-fast assertions using AgentExecution.watch().
Testing Guides
Learn how to write effective tests for your Claude Code agents using vibe-check’s powerful testing features.
Available Guides
Section titled “Available Guides”Cumulative State
Track state across multiple runs to build comprehensive test scenarios.
Custom Matchers
Use specialized matchers for files, tools, quality checks, and cost constraints.
Matrix Testing
Generate Cartesian product tests to benchmark models and configurations.
Overview
Section titled “Overview”Testing Claude Code agents requires specialized tools and patterns. These guides cover:
- Real-time monitoring - Watch execution progress and fail fast on errors
- Multi-run tracking - Aggregate state across multiple agent invocations
- Assertion patterns - Use custom matchers for common validation scenarios
- Benchmarking - Compare performance across models and configurations
Guide Details
Section titled “Guide Details”Monitor agent execution in real-time and implement fail-fast assertions.
You’ll learn:
- Using
AgentExecution.watch()for real-time monitoring - Accessing partial results during execution
- Implementing early termination conditions
- Detecting errors and anomalies as they happen
Use cases:
- Failing fast on file deletions
- Monitoring tool usage patterns
- Budget enforcement during execution
- Real-time quality gates
Track and aggregate state across multiple agent runs for comprehensive testing.
You’ll learn:
- Building cumulative context objects
- Aggregating results from multiple runs
- Cross-run state analysis
- Persistent state management
Use cases:
- Multi-step test workflows
- Progressive refinement testing
- Long-running automation validation
- State evolution tracking
Master all available matchers for comprehensive test assertions.
You’ll learn:
- File-based matchers (
toHaveChangedFiles,toHaveNoDeletedFiles) - Tool-based matchers (
toHaveUsedTool,toUseOnlyTools) - Quality matchers (
toCompleteAllTodos,toPassRubric) - Cost matchers (
toStayUnderCost) - Chaining matchers for complex assertions
Use cases:
- Validating file operations
- Enforcing tool usage policies
- Quality gates with LLM judges
- Budget enforcement
Generate Cartesian product tests for systematic model and configuration benchmarking.
You’ll learn:
- Using
defineTestSuitefor test generation - Creating configuration matrices
- Comparing model performance
- Analyzing benchmark results
Use cases:
- Model comparison (Sonnet vs Opus vs Haiku)
- Configuration optimization
- Regression testing across versions
- Performance profiling
Related Documentation
Section titled “Related Documentation”Tutorials
Section titled “Tutorials”- Your First Test - Step-by-step test creation
- Your First Evaluation - Benchmarking basics
API Reference
Section titled “API Reference”- vibeTest - Test function API
- RunResult - Result inspection interface
- Custom Matchers - Complete matcher reference
Concepts
Section titled “Concepts”- Auto-Capture - How context is captured automatically
- Lazy Loading - Memory-efficient file access