huang-sh/deepscience/evidence-synthesis

Evidence Synthesis

Use this skill for literature-backed scientific answers, research comparisons, and state-of-evidence summaries.

Workflow

1. Convert the question into answerable subquestions, populations or systems, interventions or exposures, outcomes, and relevant time constraints.
2. Define inclusion and exclusion criteria before collecting evidence.
3. Prefer primary studies and authoritative databases. Use reviews to map the field, not as substitutes for checking decisive primary claims.
4. Record each source with its identifier, study design, sample or dataset, main result, limitations, and the exact claim it supports.
5. Group evidence by claim rather than summarizing papers one at a time.
6. Explain disagreements using study design, population, measurement, model assumptions, statistical power, and publication date.
7. Assign calibrated confidence to each conclusion: high, moderate, low, or unresolved, with a short justification.
8. Identify missing evidence and the experiment or analysis most likely to resolve it.

Output Structure

• Scope and selection criteria
• Claim-to-evidence matrix
• Areas of agreement and disagreement
• Limitations and bias risks
• Calibrated conclusions
• Unresolved questions

Do not invent citations or treat a search result snippet as verified evidence.