Skip to content
Skills Library
Science & research · Analysis

bio-logic

/bio-logic

Assess a scientific claim, study design, or interpretation against its evidence. Use when testing causal reasoning, finding bias, weighing alternative explanations, or revising hypotheses.

Licence
MIT
Last repository change
01.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
8
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 01.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Assess a scientific claim, study design, or interpretation against its evidence. Use when testing causal reasoning, finding bias, weighing alternative explanations, or revising hypotheses.

Original author description

Assess a scientific claim, study design, or interpretation against its evidence. Use when testing causal reasoning, finding bias, weighing alternative explanations, or revising hypotheses.

How it works

  1. Instructions
  2. Quick Reference
  3. Input Requirements
  4. Output
  5. Quality Gates
  6. Examples
  7. Troubleshooting
  8. Related Skills
Original SKILL.md headings · 227 lines, 1,354 words
SKILL.md preview
---
name: bio-logic
description: Assess a scientific claim, study design, or interpretation against its evidence. Use when testing causal reasoning, finding bias, weighing alternative explanations, or revising hypotheses.
---

# Bio-Logic: Scientific Reasoning Evaluation

Use structured frameworks to evaluate scientific claims, methodology, and evidence strength.

## Instructions

1. Identify the task (claim assessment, paper critique, study design review, project interpretation, or hypothesis revision).
2. For project work, maintain a hypothesis register with at least 5 distinct active hypotheses until the project is no longer exploratory.
3. After each major intermediate result, reflect on what changed, revise hypothesis status, and identify the next discriminating check.
4. When findings need context, pair this reasoning with a literature-search skill such as `/polars-dovmed`, then revise hypotheses against the literature.
5. Select the study-type profile before applying a checklist. Do not grade a
   computational benchmark, phylogenetic analysis, or evolutionary inference
   as if it were an intervention trial.
6. Structure output using the provided format.

### Study-Type Profiles

| Profile | Main evidence checks | Rating language |
|---|---|---|
| Intervention | allocation, controls, adherence, attrition, estimand, harms | GRADE when the review question and evidence synthesis support it |
| Observational or quasi-experimental | identification assumptions, temporality, confounding, negative controls, sensitivity analyses | risk-of-bias plus causal-confidence statement |
| Computational or machine learning | data provenance, leakage, splits, baselines, calibration, external validation, reproducibility | computational evidence confidence |
| Evolutionary or comparative genomics | orthology, taxon/reference sampling, model fit, support, contamination, topology sensitivity | phylogenetic/comparative evidence confidence |
| Descriptive or exploratory omics | sampling, measurement, multiple testing, effect sizes, replication, alternative explanations | descriptive confidence; avoid causal labels |

### Project Hypothesis Loop

Use this loop for omics projects, unexpected results, exploratory analyses, and any request that asks what results mean.

1. **Initialize**: create at least 5 working hypotheses. Include biological mechanisms, technical artifacts, null explanations, sampling/batch effects, and annotation/database artifacts where relevant.
2. **Build an analysis playbook**: search the literature for the inferred organism, virus group, data type, or closest lineage. Summarize typical analyses, comparison baselines, markers/features, plots, and outlier criteria used by scientists in that literature.
3. **Reflect**: for every major intermediate result or QC gate, state the observation, QC status, strongest interpretation, remaining alternatives, and what evidence would separate them.
4. **Contextualize**: compare findings against the playbook and run additional literature searches for central or unexpected findings. Use broad synonym-aware queries and cite DOI/PMCID when available.
5. **Revise**: update each hypothesis as supported, weakened, ruled out, or unresolved. Keep ruled-out hypotheses visible with the evidence that changed their status.
6. **Replace**: if fewer than 5 active hypotheses remain during exploratory work, add plausible replacements or explicitly state why no additional plausible alternatives exist.

### Critique Checklist

Use relevant sections based on the review scope. Skip items not applicable to the study type.

```
## Methodology
- [ ] Design identifies the causal estimand through randomization or a justified quasi-experimental or natural-experiment strategy
- [ ] Sample size justified (power analysis reported)
- [ ] Randomization/blinding implemented where feasible
- [ ] Confounders identified and controlled
- [ ] Measurements validated and reliable

## Statistics
- [ ] Tests appropriate for data type
- [ ] Assumptions checked
- [ ] Multiple comparisons corrected
- [ ] Effect sizes + CIs reported (not just p-values)
- [ ] Missing data handled appropriately

## Interpretation
- [ ] Conclusions match evidence strength
- [ ] Limitations acknowledged
- [ ] Causal claims require an identified causal design: randomized evidence or
      a justified quasi-experimental/natural-experiment strategy with its
      assumptions and sensitivity checks
- [ ] No cherry-picking or overgeneralization

## Red Flags
- [ ] P-values clustered just below .05
- [ ] Outcomes differ from registration
- [ ] Correlation presented as causation
- [ ] Subgroups analyzed without preregistration
```

### Claim Assessment

1. Identify claim type (causal, associational, descriptive).
2. Match evidence to claim type.
3. Check logical connection between data and conclusion.
4. Ensure confidence matches evidence strength.

**Claim strength ladder**:
| Language | Requires |
|----------|----------|
| "Demonstrates" | Strong evidence under a design that identifies the target claim |
| "Suggests" / "Indicates" | Observational with controlled confounds |
| "Associated with" | Observational, no causal claim |
| "May" / "Might" | Preliminary or hypothesis-generating |

### Output Format

```markdown
## Summary
[1-2 sentences: What was studied and main finding]

## Strengths
- [Specific methodological strengths]

## Concerns
### Critical (threaten main conclusions)
- [Issue + why it matters]

### Important (affect interpretation)
- [Issue + why it matters]

### Minor (worth noting)
- [Issue]

## Evidence Rating
[Framework appropriate to the study profile, rating, and justification. Use
GRADE only when it applies to the review question.]

## Bottom Line
[What can/cannot be concluded from this evidence]
```

### Project Reasoning Output Format

```markdown
## Current Result
[Observed intermediate/final result and QC status]

## Hypothesis Register
| Rank | Hypothesis | Status | Evidence For | Evidence Against | Next Discriminating Check |
|------|------------|--------|--------------|------------------|---------------------------|
| 1 | [Hypothesis] | supported/weakened/ruled out/unresolved | [Evidence] | [Evidence] | [Test] |
| 2 | [Hypothesis] | ... | ... | ... | ... |
| 3 | [Hypothesis] | ... | ... | ... | ... |
| 4 | [Hypothesis] | ... | ... | ... | ... |
| 5 | [Hypothesis] | ... | ... | ... | ... |

## Literature Context
- [Paper or search result, DOI/PMCID if available]: supports/contradicts/narrows [hypothesis] because [reason].

## Analysis Playbook
- [Literature-derived analysis/comparison/outlier check and why it applies]

## Revision
[What changed since the previous register and why]

## Next Step
[Most informative analysis, control, or experiment]
```

## Quick Reference

| Task | Action |
|------|--------|
| Review a paper | Use the Critique Checklist above |
| Evaluate a claim | Use the Claim Assessment steps above |
| Assess evidence strength | See [references/evidence.md](references/evidence.md) |
| Identify biases | See [references/biases.md](references/biases.md) |
| Statistical errors | See [references/stats.md](references/stats.md) |
| Logical fallacies | See [references/fallacies.md](references/fallacies.md) |
| Study design review | See [references/design.md](references/design.md) |

## Input Requirements

- Manuscript or study summary
- Claim statements to evaluate
- Methods and results (as available)

## Output

- Structured critique with strengths, concerns, and evidence rating
- Clear recommendation on claim strength and limitations

## Quality Gates

- [ ] Claim type matches evidence type
- [ ] Key methodological risks are identified
- [ ] Evidence rating is justified with concrete issues
- [ ] The evidence framework matches the study type; computational and
      evolutionary work is not forced into intervention-centric GRADE rules
- [ ] Exploratory project work includes at least 5 active hypotheses or explains why fewer are plausible
- [ ] Discovery work includes a literature-derived analysis playbook before selecting analyses
- [ ] Intermediate results update hypothesis status before the next workflow step
- [ ] Central or unexpected findings are checked against the literature before final interpretation

## Examples

### Example 1: Quick Paper Critique

**Input**: "Review this RCT on vitamin D and depression"

**Output**:
```
## Summary
RCT of 200 adults testing vitamin D supplementation for depression over 12 weeks.

## Strengths
- Randomized, double-blind, placebo-controlled
- Validated depression scale (PHQ-9)
- Adequate power analysis

## Concerns
### Critical
- 35% dropout, differential by group (attrition bias)
- ITT analysis not performed

### Important
- Single-site limits generalizability

## Evidence Rating
Moderate (downgraded from high due to attrition bias)

## Bottom Line
Suggestive but not conclusive due to differential attrition.
```

### Example 2: Claim Assessment

**Input**: "This study proves that coffee prevents Alzheimer's"

**Assessment**: If based only on an association with no identified causal design, this is a correlation→causation error. Appropriate language: "Coffee consumption is associated with lower Alzheimer's risk." Observational data with a justified quasi-experimental or natural-experiment strategy, its assumptions stated and tested, can support a causal claim; the design is what decides, not whether the data were observational.

## Troubleshooting

**Issue**: Insufficient methods details
**Solution**: Request missing design/statistics information before rating evidence.

**Issue**: Conflicting results across studies
**Solution**: Report uncertainty and suggest stronger study designs for resolution.

## Related Skills

- `/ai-scientist-evaluator`: rubric-scored evaluation of AI scientist outputs
- `/manuscript-review-council`: multi-reviewer critique of peer-reviewed manuscripts
- `/proposal-review`: structured critique of funding proposals
Files in this package (8)
  • SKILL.md 10 KB
  • eval.yaml 4 KB
  • fixtures/study_summary.md 1 KB
  • references/biases.md 3 KB
  • references/design.md 3 KB
  • references/evidence.md 4 KB
  • references/fallacies.md 3 KB
  • references/stats.md 4 KB

Versions

The database has no other skills named bio-logic.

Database snapshot: 13.11.2026 · package version 2e1da4fa92ab (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100hypothesis-generation
    Formulates evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Used when turning observations or preliminary findings into…
  • 100hypothesis-generation
    Formulates evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Used when turning observations or preliminary findings into…
  • 90s4h-investigation-counter-hypothesis
    Generates the best alternative explanations for the same observations. Use when you have a claim or explanation and want to know whether something else could explain the evidence just as well — 'could there be another explanation', 'what else could cause this', 'am I sure this…
  • 100hypothesis-generation
    Formulate evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Use when turning observations or preliminary findings into transparent,…
  • 90scienceclaw-generation
    Generate scientific hypotheses, experimental designs, and paper drafts. Use when: user asks to propose hypotheses, design experiments, or write scientific content. NOT for: data analysis or literature search.
  • 100hypothesis-generation
    Formulates evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready analysis plans. Used when turning observations or preliminary findings into…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.