Skip to content

Add BLADE analysis skill for benchmark report methodology #1585

Description

@jmabry

Summary

Add a new BLADE analysis skill so agents can turn NeMo Gym rollout artifacts into evidence-backed benchmark reports and improvement recommendations.

Scope

  • Add the skill for Claude and Codex agents.
  • Include original CVDP as an opt-in, non-default example for report interpretation.
  • Reference existing CVDP example rollouts in resources_servers/cvdp/data/ instead of duplicating rollout JSONL.
  • Include only Nemotron 3 Super golden analysis artifacts as optional reference material.
  • Keep the skill and artifacts sanitized for public repo use.

Draft PR

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions