tooluniverse-self-review

Installation
SKILL.md

Self-Review: Task Success Criteria + Work Review

Derive the criteria that define a good result for a specific task, then review work against them. Each task defines its own "evaluation world" -- a set of scenarios, perspectives, and criteria that capture what matters for judging results for that specific task. Built on the Qworld Recursive Expansion Tree (RET).

When to Use

  • An agent (or person) is about to do, is doing, or has finished a task and wants to know what a good result must cover, or what is missing or weak.
  • You need a definition-of-done or completeness checklist for a task or goal.
  • You need an evaluation rubric / grading criteria for answers to an open-ended question.
  • You want to score or grade responses: LLM-as-judge, exam answers, candidate responses.

Inputs

Installs
25
GitHub Stars
1.5K
First Seen
Jun 21, 2026
tooluniverse-self-review — mims-harvard/tooluniverse