scientific-research-design-method

# Scientific research design method ## Upstream provenance This is a bounded OpenCorvus adaptation of the individually MIT-licensed `scientific-brainstorming` Skill from [K-Dense-AI/scientific-agent-skills](https://github.com/K-Dense-AI/scientific-agent-skills), pinned to `7eb9c23c32ecf7f8c19cb45ded3150534ccefe6a`. See `references/upstream.md` and `references/upstream-license.txt`. It retains independent idea generation, explicit assumptions, adversarial review, and decision logging while excluding upstream CLIs, cross-Skill dependencies, and field-specific operational instructions. ## Workflow 1. Freeze one research question, decision owner, evidence date, constraints, prohibited outputs, and required institutional reviews. 2. Separate located evidence, ideas, assumptions, predictions, unknowns, and decisions. 3. Generate competing hypotheses independently before discussion or scoring. 4. Record disconfirming observations, alternatives, confounders, information gain, feasibility, and null-result value. 5. Run rigor, ethics, safety, privacy, equity, and dual-use review; a score never overrides a gate. 6. Join into a decision register with dissent, source limits, proposed tests, approvals, and revisit triggers. ## Boundaries - Do not provide patient-specific advice, claim empirical validation, or approve regulated research. - Stop and route harmful, evasive, clinical, biosafety, or controlled work to authorized professionals. - Use `assets/research-decision-register.md` for the final deliverable.