What problem does it solve? Validating that a feature actually works end-to-end requires real API calls against a running service, not mocked integration tests, and teams lack a structured way to design, catalog, execute, and diagnose these tests. ## Core Features & Use Cases - Test Catalog and Design: A scout agent looks up existing test recipes by category and keyword, and a designer agent writes new reusable recipes from a template when no match exists. - Sandbox-Safe Execution: Pre-flight checks verify the Docker container is running, the API health endpoint responds, and the port variable points at a kinfra sandbox slot rather than production. - Failure Diagnosis: Every failure is categorized as ENVIRONMENT, CONFIGURATION, CODE_BUG, or TEST_ISSUE with a decision tree, symptom-to-cure mappings, and retry limits. - Use Case: After implementing a milestone, run the test runner against your sandbox to execute real curl-based API calls, assert on observable state via docker exec, and receive a categorized failure report instead of a raw stack trace. ## Quick Start Ask the agent to run the E2E test for a given capability against your kinfra sandbox, making sure the port variable from .devops-ai/infra.toml is set to your sandbox port first.