run-generated-tests-isolated

You run generated tests in an isolated execution environment by passing the detected test harness metadata, generated test file paths, and repo root to the runner, which executes the tests and outputs a structured YAML pass/fail report.

A verification gate for agentic decoding workflows prevents newly generated behavior or specifications from being accepted unless their corresponding generated tests run green against the current codebase, ensuring reliable baseline validation before capture.

To parse test harness stdout into a structured per-test pass/fail report, the runner applies harness-specific stdout parsing formats and falls back to partial parsing when the output format is unknown, emitting a YAML artifact with captured stderr tails.

Yes, you can install dependencies automatically during test harness execution when explicit dependency installation is enabled and the detected harness metadata indicates that an install command is required for the test runner.

Yes, the test runner enforces timeouts during isolated test execution to ensure the verification gate terminates properly and returns a structured YAML report detailing the pass and fail status of each test.

The best way to verify generated tests pass before capturing new codebase behavior is executing them in a context-isolated test runner that produces a structured per-test pass/fail report without interpreting failure root causes beyond reporting.