artifact-create

This Skill packages research code into a reproducible artifact by cleaning the project, baking a pinned Docker environment, and verifying results in a clean-room container to ensure evaluators can reproduce your claimed findings end-to-end.

Clean-room verification builds your Docker image and reproduces key results from scratch in an isolated container, proving the artifact is self-contained and evaluator-ready for ACM or USENIX reproducibility badges.

Preparing an Artifact Appendix and result-to-command map is handled by this Skill, which generates the necessary AE paperwork including README, INSTALL files, and access plans alongside your packaged Docker artifact.

Yes, this Skill uses Docker with fully pinned dependencies and fixed base images to guarantee reproducibility, handling large data via mounts or checksummed downloads while avoiding baked secrets.

Yes, the packaging process uses a clean-first, behavior-preserving approach that runs code-clean and verifies the cleaned code still reproduces the exact same claimed results before finalizing the Docker artifact.

Limitations include avoiding baked secrets and preventing the use of latest tags for base images; the process requires fixed base images, fully pinned dependencies, and explicit handling of large data via mounts or checksummed downloads.