This project contains minimal, fast-feedback integration tests for OpenExec. It is designed to verify the core orchestration loop, context assembly, and blueprint execution in under 60 seconds per scenario.
01-smoke-web: Verifies basic HTML file modification and CSS injection.02-broken-logic: Verifies agentic debugging. Fixes a Python subtraction bug and runspytestto confirm success.
The harness can automatically build and verify the OpenExec version before running scenarios.
Run all scenarios with auto-build:
./run-harness.sh --buildRun all scenarios using existing binary (with version check):
./run-harness.shRun a single scenario:
./run-harness.sh 01-smoke-web- Version Integrity: Automatically warns if your testing binary is out of sync with your source code.
- Prevent 80% Plateaus: Ensures that code changes aren't just "scaffolded" but actually implemented.
- Fast Iteration: Test prompt and blueprint changes in seconds, not minutes.
- Deterministic Control: Uses the local OpenExec binary to prove that core fixes (permissions, task descriptions) work in practice.