review-research
О программе
Этот навык обеспечивает структурированное рецензирование исследовательских материалов через оценку методологии, статистической корректности и воспроизводимости. Используйте его для проверки рукописей, препринтов, исследовательских заявок или глав диссертаций на научную строгость и предвзятость. Он предоставляет конструктивную обратную связь, анализируя дизайн эксперимента и качество доказательств.
Быстрая установка
Claude Code
Рекомендуетсяnpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/review-researchСкопируйте и вставьте эту команду в Claude Code для установки этого навыка
Документация
Review Research
Structured peer review of research, eval methodology, statistical choices, reproducibility, overall scientific rigour.
Use When
- Review manuscript, preprint, internal research report
- Eval research proposal or study protocol
- Assess evidence quality behind claim/recommendation
- Provide feedback on colleague's design before data collection
- Review thesis chapter or dissertation section
In
- Required: Research doc (manuscript, report, proposal, protocol)
- Required: Field/discipline ctx (affects methodology stds)
- Optional: Journal/venue guidelines (if reviewing for publication)
- Optional: Supplementary materials (data, code, appendices)
- Optional: Prior reviewer comments (if reviewing revision)
Do
Step 1: First Pass — Scope + Structure
Read entire doc once → understand:
- Research q: Clearly stated + specific?
- Contribution claim: What is novel?
- Overall structure: Follows expected format (IMRaD, venue-specific)?
- Scope match: Appropriate for target audience/venue?
## First Pass Assessment
- **Research question**: [Clear / Vague / Missing]
- **Novelty claim**: [Stated and supported / Overstated / Unclear]
- **Structure**: [Complete / Missing sections: ___]
- **Scope fit**: [Appropriate / Marginal / Not appropriate]
- **Recommendation after first pass**: [Continue review / Major concerns to flag early]
→ Clear understanding of paper's claims + contribution. If err: research q unclear after full read → note as major concern + proceed.
Step 2: Eval Methodology
Assess design vs std for field:
Quantitative Research
- Study design appropriate for q (experimental, quasi-experimental, observational, survey)
- Sample size justified (power analysis or practical rationale)
- Sampling method described + appropriate (random, stratified, convenience)
- Vars clearly defined (independent, dependent, control, confounding)
- Measurement instruments validated + reliability reported
- Data collection procedure reproducible from description
- Ethical considerations addressed (IRB/ethics approval, consent)
Qualitative Research
- Methodology explicit (grounded theory, phenomenology, case study, ethnography)
- Participant selection criteria + saturation discussed
- Data collection methods described (interviews, observations, documents)
- Researcher positionality acknowledged
- Trustworthiness strategies reported (triangulation, member checking, audit trail)
- Ethical considerations addressed
Mixed Methods
- Rationale for mixed design explained
- Integration strategy described (convergent, explanatory sequential, exploratory sequential)
- Both quant + qual components meet respective stds
→ Methodology checklist completed w/ specific obs each item. If err: critical methodology missing → flag major concern not assume.
Step 3: Statistical + Analytical Choices
- Stats appropriate for data type + research q
- Assumptions checked + reported (normality, homoscedasticity, independence)
- Effect sizes reported alongside p-values
- Confidence intervals provided where appropriate
- Multi comparison corrections applied when needed (Bonferroni, FDR, etc.)
- Missing data handling described + appropriate
- Sensitivity analyses for key assumptions
- Results interpretation consistent w/ analysis (no overstating)
Common stat red flags:
- p-hacking indicators (many comparisons, selective reporting, "marginally significant")
- Inappropriate tests (t-test on non-normal w/o justification, parametric on ordinal)
- Confusing statistical significance w/ practical
- No effect size reporting
- Post-hoc hypotheses presented as a priori
→ Stat choices eval'd w/ specific concerns documented. If err: reviewer lacks expertise in specific method → ack + recommend specialist reviewer.
Step 4: Reproducibility
- Data availability stated (open data, repo link, on req)
- Analysis code availability stated
- Software vers + envs documented
- Random seeds or reproducibility mechanisms described
- Key params + hyperparameters reported
- Computational env described (hardware, OS, deps)
Reproducibility tiers:
| Tier | Description | Evidence |
|---|---|---|
| Gold | Fully reproducible | Open data + open code + containerized environment |
| Silver | Substantially reproducible | Data available, analysis described in detail |
| Bronze | Potentially reproducible | Methods described but no data/code sharing |
| Opaque | Not reproducible | Insufficient method detail or proprietary data |
→ Reproducibility tier assigned w/ justification. If err: data can't be shared (privacy, proprietary) → synthetic data or detailed pseudocode acceptable alt — note if provided.
Step 5: Identify Biases
- Selection bias: Participants representative of target pop?
- Measurement bias: Could measurement systematically distort results?
- Reporting bias: All outcomes reported, including non-significant?
- Confirmation bias: Did authors only look for evidence supporting hypothesis?
- Survivorship bias: Dropouts, excluded data, failed exps accounted for?
- Funding bias: Funding source disclosed + could influence findings?
- Publication bias: Complete picture or might negative results be missing?
→ Potential biases ID'd w/ specific examples from manuscript. If err: biases can't be assessed from available info → recommend authors address explicitly.
Step 6: Write Review
Structure constructive:
## Summary
[2-3 sentences summarizing the paper's contribution and your overall assessment]
## Major Concerns
[Issues that must be addressed before the work can be considered sound]
1. **[Concern title]**: [Specific description with reference to section/page/figure]
- *Suggestion*: [How the authors might address this]
2. ...
## Minor Concerns
[Issues that improve quality but are not fundamental]
1. **[Concern title]**: [Specific description]
- *Suggestion*: [Recommended change]
## Questions for the Authors
[Clarifications needed to complete the evaluation]
1. ...
## Positive Observations
[Specific strengths worth acknowledging]
1. ...
## Recommendation
[Accept / Minor revision / Major revision / Reject]
[Brief rationale for the recommendation]
→ Review specific, constructive, refs exact locations in manuscript. If err: review running long → prioritize major concerns + note minor in summary list.
Check
- Every major concern refs specific section, figure, claim
- Feedback constructive — problems paired w/ suggestions
- Positive aspects ack'd alongside concerns
- Stat assessment matches analysis methods used
- Reproducibility explicitly eval'd
- Recommendation consistent w/ severity of concerns raised
- Tone professional, respectful, collegial
Traps
- Vague criticism: "Methodology is weak" unhelpful. Specify what + why.
- Demand diff study: Review research done, not research you would have done.
- Ignore scope: Conference paper has diff expectations than journal article.
- Ad hominem: Review work, not authors. Never ref author identity.
- Perfectionism: No study perfect. Focus on concerns changing conclusions.
→
review-data-analysis— deeper focus on data quality + model validationformat-apa-report— APA formatting stds for research reportsgenerate-statistical-tables— publication-ready statistical tablesvalidate-statistical-output— statistical output verification
GitHub репозиторий
Похожие навыки
executing-plans
ДизайнИспользуйте навык executing-plans, когда у вас есть полный план реализации для выполнения контролируемыми партиями с контрольными точками проверки. Он загружает и критически анализирует план, затем выполняет задачи небольшими партиями (по умолчанию 3 задачи), сообщая о прогрессе между каждой партией для проверки архитектором. Это обеспечивает систематическую реализацию со встроенными контрольными точками проверки качества.
requesting-code-review
ДизайнЭтот навык запускает суб-агента для ревью кода, который анализирует изменения в коде на соответствие требованиям перед дальнейшими действиями. Его следует использовать после завершения задач, реализации крупных функций или перед слиянием с основной веткой. Ревью помогает выявить проблемы на ранней стадии, сравнивая текущую реализацию с исходным планом.
connect-mcp-server
ДизайнЭтот навык предоставляет разработчикам подробное руководство по подключению серверов MCP к Claude Code с использованием транспортов HTTP, stdio или SSE. Он охватывает установку, конфигурацию, аутентификацию и безопасность для интеграции внешних сервисов, таких как GitHub, Notion и пользовательские API. Используйте его при настройке интеграций MCP, конфигурации внешних инструментов или работе с Model Context Protocol от Claude.
web-cli-teleport
ДизайнЭтот навык помогает разработчикам выбирать между веб-интерфейсом Claude Code и CLI на основе анализа задачи, а также обеспечивает бесшовное перемещение сессий между этими средами. Он оптимизирует рабочий процесс, управляя состоянием и контекстом сессии при переключении между веб-интерфейсом, CLI или мобильным приложением. Используйте его для сложных проектов, требующих различных инструментов на разных этапах работы.
