scientific-brainstorming

Define one focal question, have participants generate ideas independently and privately before any discussion, then share round-robin, cluster by an explicit relation, define criteria before scoring, and run adversarial review. The skill's SKILL.md provides a ten-step reproducible workflow.

Run python scripts/session_scaffold.py with --session-id, --title, --question, and repeated --participant flags, plus an --output path. It generates a deterministic JSON register with workflow stages, templates, and governance fields.

Define criteria and weights in a JSON config, record ratings in a CSV with optional low/high uncertainty columns, then run scripts/evaluate_matrix.py. It outputs normalized scores, intervals, one-at-a-time weight sensitivity, and leaves the decision field null.

No. The bundled CLIs require only Python 3.11+ and the standard library. They make no network or LLM calls, need no credentials, and refuse symlinks and unbounded inputs for safety.

No. Brainstorming produces proposals, not findings. Hypothesis validation requires observations, suitable designs, and independent scrutiny; the skill routes empirical validation, ethics review, and clinical questions to appropriate experts or other skills.

The additive model allows compensation across criteria, so hard ethics or feasibility gates must stay outside the score. Sensitivity analysis is local and one-factor-at-a-time, and ratings remain judgments rather than empirical measurements.