Scientific Brainstorming Facilitator

What it does

Structures research brainstorming so that ideas, assumptions, evidence, and decisions never get blurred together:

• Scoping: one focal question plus decision owner, constraints, out-of-scope topics, prohibited outputs, and whether human subjects, animals, sensitive data, pathogens, or dual-use tech are implicated
• Independent-first generation: private, parallel idea writing before anyone sees peer or AI examples (avoids production blocking and anchoring)
• Provenance capture: every idea gets an ID, contributor, stage (independent / discussion / post-check), and origin (human, AI-assisted, literature-inspired, mixed)
• Transparent evaluation: criteria, scale anchors, and weights defined before scoring; raw ratings, uncertainty ranges, dissent, and abstentions stay visible
• Adversarial review: a non-originator asks what would falsify the idea, what alternative explanation fits, what confounder or measurement failure could dominate
• Gates and decision log: feasibility, rigor, ethics, biosafety, and dual-use gates that a high creativity score can never override
• Optional local CLIs: session_scaffold.py, validate_register.py, evaluate_matrix.py — deterministic, standard-library only, no network or LLM calls

It explicitly refuses to validate hypotheses, and hands off to hypothesis-generation, experimental-design, statistical-power, literature-review, and statistical-analysis.

Who it's for

• Grad students, postdocs, and PIs choosing and prioritizing research directions
• Facilitators running lab meetings or research workshops who want less bias and a paper trail
• Grant-writing teams narrowing many candidate aims with documented reasoning
• R&D groups that need an auditable record of how an idea was selected

Example uses

1. Run a lab meeting: "Design a brainstorming session on mechanisms linking gut microbiome composition to metabolic phenotype." → returns a neutral prompt, timed private-writing round, round-robin sharing script, and clustering rules.
2. Prioritize a candidate list: give it 12 ideas and your criteria; it builds a weighted matrix (information gain, originality vs. checked literature, feasibility, ethics burden), runs one-at-a-time weight sensitivity via evaluate_matrix.py, and returns a shortlist with decision deliberately left null.
3. Red-team a favorite hypothesis: it generates the falsification questions, alternative explanations, hidden-dependency checks, and a template recording mitigation plus residual uncertainty rather than a pass/fail stamp.

Install the skill I found on Claude Skill Mart.
Copy the skills/scientific-brainstorming folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/scientific-brainstorming/.
When it's done, tell me in one line what this skill can do.