name	refined-idea-quality
description	Quality criteria for refined_idea output. Use before RefineIdea or when ValidateRefinedIdea returns low score. Focus on concrete, decomposable, and specific output.

Refined Idea Quality Criteria

Guidelines for producing a high-quality refined idea (Markdown) that is decomposable into 3–10 tasks. Quality matters.

Concrete over vague: "Compare BERT and GPT on code completion benchmarks" not "Research NLP models".
Combine literature insights with novel angles: Avoid mere summarization; add synthesis, comparison, or gap.
Must be decomposable: If you cannot list 3+ distinct sub-tasks, refine further.
Gap must be specific: Name limitations (e.g. "prior work focuses on X, neglects Y") rather than generic "more research needed".

Score >= 4: Proceed to FinishIdea.
Score < 4: Call RefineIdea again with improvements — tighten scope, clarify questions, or sharpen gap.

name	refined-idea-quality
description	Quality criteria for refined_idea output. Use before RefineIdea or when ValidateRefinedIdea returns low score. Focus on concrete, decomposable, and specific output.

Guidelines for producing a high-quality refined idea (Markdown) that is decomposable into 3–10 tasks. Quality matters.

Concrete over vague: "Compare BERT and GPT on code completion benchmarks" not "Research NLP models".
Combine literature insights with novel angles: Avoid mere summarization; add synthesis, comparison, or gap.
Must be decomposable: If you cannot list 3+ distinct sub-tasks, refine further.
Gap must be specific: Name limitations (e.g. "prior work focuses on X, neglects Y") rather than generic "more research needed".

Score >= 4: Proceed to FinishIdea.
Score < 4: Call RefineIdea again with improvements — tighten scope, clarify questions, or sharpen gap.

refined-idea-quality