DidierRLopes/openbb-workspace-bench

task-author

Author complete workspace_tasks task.json files (setup, eval, reference trace, prompt) from a persona README and a story markdown. Use when creating or revising tasks for the workspace_tasks taskset.

Idioma do texto original: inglês

workspace-bench-tasks-orchestrator

Coordinate the authoring of workspace_tasks stories - spawn one task-author subagent per story (each pipelines its own per-level validators), trigger the ladder fairness check, relay flags, and stop for the user's review after each completed persona. Use when…

Idioma do texto original: inglês

task-level-fairness

Judge a completed story ladder as a set - the five task.json files of one story compared side by side, verifying each level genuinely differentiates according to the ladder rules. Triggered once all five levels of a story exist and pass per-task validation.

Idioma do texto original: inglês

task-validator

Fresh-eyes feasibility judgment of workspace_tasks task.json files - could an agent given only the prompt and the tools actually satisfy the eval? Use after tasks are authored and certified, before they count.

Idioma do texto original: inglês