What problem does it solve?
It prevents newly generated behavior/specifications from being accepted unless the corresponding generated tests run green against the current codebase, ensuring a reliable verification gate.
Core Features & Use Cases
- Context-isolated execution: Designed to be consumed only by the test-runner agent in context isolation, so the execution environment is controlled and the runner only sees the intended test files and codebase.
- Harness-driven test running: Executes the correct test command using detected
test_harness metadata (runner command, working directory, env vars, install command, and output parsing format).
- Pass/fail per test reporting: Produces a structured YAML report with totals and per-test status, including captured stderr, without interpreting failure root causes beyond reporting.
- Optional dependency installation: Can install dependencies when explicitly enabled and when the harness indicates installation is required.
Use case: During /decode, every captured behavior/flow/spec is verified by running its generated tests before it’s considered “captured,” forming the baseline-green gate for C25.
Quick Start
Ask the test-runner agent to call run-generated-tests-isolated with the detected test harness, the generated test file paths, and the repo root so it returns a pass/fail report artifact.