# Evaluate Scientific Evidence with Rigor

Scientific claims can overstate what methods and data support. This skill applies structured checks for validity, bias, statistics, reasoning, and evidence quality.

## Install

```bash
npx skillstore add davila7/scientific-critical-thinking
```

## Metadata

- Status: approved
- Slug: davila7-scientific-critical-thinking
- Skillstore revision: r2
- Version status: missing
- Tree hash: d5f391bb46e91b102504307ff30f628d1c1ca26e558546c4480817618e9d94ea
- Author: davila7
- GitHub username: davila7
- License: MIT
- Repository: https://github.com/davila7/claude-code-templates/tree/main/cli-tool/components/skills/scientific/scientific-critical-thinking
- Ref: c43861a65bb95efcae259cd161c9d6f4dc7eec6f
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 75
- Public page: https://skillstore.pages.dev/skills/davila7-scientific-critical-thinking
- Manifest: https://skillstore.pages.dev/api/skills/davila7-scientific-critical-thinking/manifest

## Capabilities

- Assesses alignment between research questions, study designs, measurements, and causal claims.
- Identifies cognitive, selection, measurement, reporting, and analysis biases.
- Reviews sample size, statistical tests, assumptions, p-values, effect sizes, confidence intervals, missing data, and multiple comparisons.
- Applies evidence hierarchy, GRADE considerations, and risk-of-bias principles.
- Recommends practical improvements to study designs and scientific claims.

## Use Cases

- Strengthen a Study Protocol: Find design, sampling, measurement, bias, and analysis weaknesses before data collection begins.
- Review a Manuscript: Separate fatal validity threats from important limitations and minor reporting concerns.
- Check a Public Science Claim: Compare claim strength with the underlying study design, evidence, and uncertainty.

## Prompt Templates

### Quick Claim Check

```
Evaluate this scientific claim: [claim]. Identify the evidence needed, likely alternative explanations, and any unsupported causal language.
```

### Study Methods Review

```
Review this study summary: [methods and results]. Assess design fit, sample quality, measurement validity, bias, confounding, and statistical interpretation. Separate major and minor concerns.
```

### Evidence Quality Comparison

```
Compare these studies: [study summaries]. Rate evidence quality using relevant GRADE considerations and risk-of-bias domains. Explain disagreements and synthesis limits.
```

### Protocol Stress Test

```
Critique this research protocol: [protocol]. Map each validity threat to its likely impact. Propose design or analysis changes, then identify unresolved uncertainty.
```

## Limitations

- Does not retrieve papers, verify citations, or access external databases without separate tools and supplied sources.
- Cannot replace domain experts, statisticians, ethics review boards, or formal systematic review procedures.
- Conclusions depend on the methodological details and evidence supplied in the prompt.
- Provides structured critique, not clinical, legal, or policy advice.

## Best Practices

- Provide the full methods, results, tables, and exact claim whenever possible.
- Separate observed data from interpretation, then match confidence to evidence strength.
- Ask for conditional conclusions when key design or analysis details are missing.

## Anti Patterns

- Do not treat one checklist item as proof that an entire study is invalid.
- Do not infer causation from association without a defensible causal design.
- Do not equate statistical significance with practical importance or evidence certainty.

## Security Audit

- Audited at: 2026-07-23T14:34:00.983\+00:00
- Summary: All 32 static findings are false positives caused by scientific prose or Markdown formatting. The command example invokes a local schematic generator and shows no download, privilege escalation, persistence, exfiltration, or prompt injection. No semantic security findings were identified.

## Stats

- Views: 273
- Downloads: 9
- Favorites: 0
- Popularity score: 0