Calibrate assessment rubrics by reviewing agent work in GitHub/GitLab PRs and feeding human comments back into the rubric. Use this skill when the user wants to: - Calibrate, tune, or sanity-check assessment criteria / dimensions of a benchmark - Review trial diffs alongside the LLM-as-a-Judge score
Calibrate assessment rubrics by reviewing agent work in GitHub/GitLab PRs and feeding human comments back into the rubric. Use this skill when the user wants to: - Calibrate, tune, or sanity-check assessment criteria / dimensions of a benchmark - Review trial diffs alongside the LLM-as-a-Judge scores in a PR/MR - Investigate why judge scores feel off, too harsh, too lenient, or misaligned with how
Cette entrée ne publie aucun paquet npm : Forge n'a donc pas d'arbre de dépendances pour elle. C'est une lacune de couverture — pas une affirmation qu'elle n'a aucune dépendance.