Mind production-harness qualification exposed a replay-unstable default Python runtime result. When an explicit pytest target such as CalculatorTests.test_subtract names no collectible test, pytest exits non-zero and emits file or directory not found. RunPythonTestsStructured currently runs generic environment classification first, producing env-blocked (unknown) with transient uv/JUnit paths. The runtime contract already states selector-scoped zero-match must be test-selection-error (no-tests-matched-selectors). Fix the default runner to recognize pytest zero-match signals before generic environment classification, return a stable typed diagnostic, preserve true environment failures, and prove it through the real uv/pytest runner.
Mind production-harness qualification exposed a replay-unstable default Python runtime result. When an explicit pytest target such as
CalculatorTests.test_subtractnames no collectible test, pytest exits non-zero and emitsfile or directory not found.RunPythonTestsStructuredcurrently runs generic environment classification first, producingenv-blocked (unknown)with transient uv/JUnit paths. The runtime contract already states selector-scoped zero-match must betest-selection-error (no-tests-matched-selectors). Fix the default runner to recognize pytest zero-match signals before generic environment classification, return a stable typed diagnostic, preserve true environment failures, and prove it through the real uv/pytest runner.