Scientific Brainstorming
Use this skill to generate research ideas that are useful, testable, and honest about uncertainty.
- Define the observation, domain, population or system, and the decision the idea should support.
- Separate established evidence, plausible interpretation, and speculation. Use the selected ScanSci knowledge scope when available; do not treat a search lead as evidence.
- Produce a compact idea card for each candidate:
- question and falsifiable hypothesis;
- mechanism or causal chain, with competing explanations;
- minimum experiment or analysis;
- expected observations under each hypothesis;
- controls, confounders, feasibility risks, and kill criteria;
- novelty claim and the evidence needed to verify it.
- Rank ideas by importance, tractability, discriminating power, and evidence readiness. Do not rank by novelty alone.
- End with the next smallest action: search, measurement, pilot, simulation, or expert review.
Never invent a literature gap, preliminary result, sample size, mechanism, or citation. Mark unresolved assumptions explicitly. When a user asks for a grant or paper angle, stabilize the research question before polishing the story.
This is a ScanSci adaptation of the public K-Dense Scientific Agent Skills scientific-brainstorming workflow. It contains no upstream scripts or dependency installer.
1---2name: scientific-brainstorming3description: Evidence-bounded scientific ideation for research questions, hypotheses, mechanisms, experiments, and falsifiers. Use for research brainstorming, study design, gap analysis, or turning observations into testable ideas.4---56# Scientific Brainstorming78Use this skill to generate research ideas that are useful, testable, and honest about uncertainty.9101. Define the observation, domain, population or system, and the decision the idea should support.112. Separate established evidence, plausible interpretation, and speculation. Use the selected ScanSci knowledge scope when available; do not treat a search lead as evidence.123. Produce a compact idea card for each candidate:13 - question and falsifiable hypothesis;14 - mechanism or causal chain, with competing explanations;15 - minimum experiment or analysis;16 - expected observations under each hypothesis;17 - controls, confounders, feasibility risks, and kill criteria;18 - novelty claim and the evidence needed to verify it.194. Rank ideas by importance, tractability, discriminating power, and evidence readiness. Do not rank by novelty alone.205. End with the next smallest action: search, measurement, pilot, simulation, or expert review.2122Never invent a literature gap, preliminary result, sample size, mechanism, or citation. Mark unresolved assumptions explicitly. When a user asks for a grant or paper angle, stabilize the research question before polishing the story.2324This is a ScanSci adaptation of the public K-Dense Scientific Agent Skills `scientific-brainstorming` workflow. It contains no upstream scripts or dependency installer.
Rimagination