domains / research / llm-eval-pipeline-audit / eval-audit

eval-audit

Skill card

authentic

The two copies are the same procedure with near-identical structure, checks, and findings. The ai-evals-course copy is the current, non-archived version and uses the current skill names (error-discovery, write-code-eval) referenced by the surrounding skill set, while the hamelsmu copy is archived and uses older names (error-analysis) and omits the write-code-eval reference. The differences are naming/version drift, not substantive alternative steps, so the current copy is the authentic procedure the other reproduces.

Repositories

RepositoryPath
ai-evals-course/evals-skillsskills/eval-audit/SKILL.md

Other copies

RoleRepositoryPath
inferiorhamelsmu/evals-skillsskills/eval-audit/SKILL.md