Self-hosted alternatives to explore-run

E

explore-run

Rigor Improve / Rigor Explore run leaf skill

explore-run is an agent skill from the rigorpilot-skills collection that plans and runs bounded exploratory experiments on a deep learning research repository. It is the Rigor Improve and Rigor Explore run leaf skill, and its installed name is kept for compatibility. It is for the case where a researcher explicitly authorizes exploratory runs, such as a small-subset validation, a short training probe, a batch sweep, a search that uses idle GPU capacity or a quick transfer-learning trial. The purpose of these runs is to produce evidence that ranks candidate runs against each other, so the skill is careful not to present an output as a certified success. It preserves the model's judgment about the repository it is working in, so the planning depends on the specific project rather than on a fixed template. The skill is meant to keep exploratory work small and bounded, which helps avoid burning compute on runs that were never well planned. It is a companion to the explore-code skill for candidate changes and the run-train skill for full training. It should not be used when the researcher has not authorized exploration, or when the goal is a trusted, final result. The skill passes Socket and Snyk, and Gen Agent Trust Hub shows a warning on skills.sh. Install it with npx skills add https://github.com/lllllllama/rigorpilot-skills --skill explore-run.