These are the harness configurations behind We mutation-tested 10 popular open-source projects across five languages. One directory per project, holding the flawd.toml, the Dockerfile.flawd the runs executed in, and any wrapper script the test command calls.
Publishing these is part of the study's methodology. No mutation-testing harness runs "the test suite," each one runs a reduction of it, and the reduction moves the numbers. The study's appendix describes each reduction in prose; these are the actual configs, so you can check that prose against reality or reproduce a run.
Notes:
- The initial survey runs used a pre-release flawd build in July 2026; verification and audit reruns used released binaries, v0.9.0 through v0.12.0. The version and upstream commit per project are recorded in the study's raw data.
- The configs are as last run. Upstream moves fast (Ghost switched both package manager and test runner mid-study), so expect drift against today's main branches.
- Wrapper scripts exist where the project's test invocation needed massaging, such as worker caps or quarantining a flaky test. The quarantines are part of the reduction and are listed in the study appendix.
Flawd itself is at fixture.dev/flawd.