Rimagination/good-question

What it solves

Researchers often struggle to move from a broad interest or a vague literature gap to a concrete, high-quality research question. This project prevents researchers from pursuing "weak" questions—those that are too broad, lack competing hypotheses, or cannot be falsified—by providing a structured framework to sharpen and stress-test ideas before investing significant time and resources.

How it works

It is a portable agent skill (compatible with Codex and Claude Code) that implements a strict scientific workflow. The process involves:

1. Diagnosis: Identifying the user's current state (e.g., broad interest, existing proposal, or stalled project).
2. Mode Selection: Switching between roles like mentor, reviewer, or collaborator based on the context.
3. Contextualization: Building domain briefs and loading field-specific adapters (e.g., for AI4Science or biomedicine) to ensure grounding in evidence.
4. Refinement: Using structured lenses to generate candidates, then converging them based on importance, feasibility, and falsifiability.
5. Validation: Running a final "editor-desk reject" gate to ensure the question is robust enough to withstand professional review.

Who it’s for

Academic researchers and students across various fields—including ecology, remote sensing, AI4Science, social sciences, and biomedicine—who need to refine their research directions, write grant proposals, or conduct rebuttals.

Highlights

• Good Question Cards: Produces structured outputs covering working titles, core assumptions, competing hypotheses, and two-week pilot plans.
• Multi-Role Simulation: Simulates different perspectives (mentor vs. reviewer) to identify potential objections and repair weak points.
• **Methodologically Grounded: Based on established scientific thinking frameworks such as the Heilmeier Catechism and Strong Inference.
• Source Auditing: Includes a mechanism to verify if literature gaps and contribution claims are actually supported by public sources.