run-tests

To run benchmark tests in Docker, this Skill automates eval-snapshot execution against a given snapshot, problem name, and checkpoint index to produce structured evaluation results. It ensures isolation and reproducible results for problem evaluations.

Evaluating problem checkpoints generates evaluation.json, evaluation.log, and a quality_analysis package. These outputs provide structured benchmark results and code quality metrics to a specified directory for CI pipelines.

Yes, this Skill uses eval-snapshot instead of raw pytest to ensure isolation and reproducible results. It is ideal for evaluating problem checkpoints for slop-code-bench and similar projects in Docker.

Yes, Docker is required to execute benchmark tests with eval-snapshot. The Skill automates testing of coding solutions inside a controlled Docker environment for both CI pipelines and local experimentation.

Evaluating a problem checkpoint requires a path to a snapshot, a problem name, and a checkpoint index. These inputs allow the evaluator to cross-check multiple checkpoints across problems in SlopCodeBench.