# Evidence Review

> Independently review a life-science manuscript, analysis, figure package, or repository for claim support, citation validity, numeric traceability, figure-code-data consistency, reproducibility, and privacy. Use as a final gate, not as author self-approval.

- Skill: `minimax-ai/evidence-review` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx get-skill add minimax-ai/evidence-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/minimax-ai/evidence-review/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: MiniMax AI (https://skillmd.com/u/minimax-ai)
- Updated: 2026-11-11
- Page: https://skillmd.com/skills/minimax-ai/evidence-review

---


# Evidence Review

Read [references/review-contract.md](references/review-contract.md). Review the artifacts and acceptance criteria, not the producer's confidence or narrative.

## Review layers

1. `Scientific`: question, design, controls, experimental unit, and causal strength.
2. `Citation`: existence, bibliographic correctness, primary-source preference, and semantic support for the exact clause.
3. `Numbers`: every sample size, percentage, effect, interval, p value, and identifier traced to source data or output.
4. `Figures`: source table, code, labels, statistics, legends, and manuscript text agree.
5. `Reproducibility`: clean rerun, versions, parameters, seeds, failures, and environment evidence.
6. `Privacy`: no private dataset, credential, local path, or excluded organism-specific resource entered a reusable package.

## Independence rule

If the reviewer participated in producing an artifact, disclose that conflict and perform only a provisional check. For a consequential gate, use a fresh review context or independent agent when available.

## Verdict

Return exactly one: `PASS`, `PASS WITH REQUIRED REVISIONS`, or `BLOCK`. A PASS requires evidence for each required layer; silence is not evidence.

