sciagent-skill-creator
Scaffold a new SciAgent-Skills entry. Picks pipeline/toolkit/database/guide template, creates skills/{category}/{name}/SKILL.md with valid frontmatter, appends…
Structured peer review of manuscripts and grants. 7-stage evaluation: initial assessment, section review, statistical rigor, reproducibility, figure integrity, ethics, writing. Covers CONSORT/STROBE/PRISMA and report structure. For evidence quality see
$ npx -y skills add jaechang-hits/SciAgent-Skills --skill peer-review-methodology --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/peer-review-methodologyContext preview
The summary Claude sees to decide when to auto-load this skill.
Structured peer review of manuscripts and grants. 7-stage evaluation: initial assessment, section review, statistical rigor, reproducibility, figure integrity, ethics, writing. Covers CONSORT/STROBE/PRISMA and report structure. For evidence quality see
name: peer-review-methodology description: "Structured peer review of manuscripts and grants. 7-stage evaluation: initial assessment, section review, statistical rigor, reproducibility, figure integrity, ethics, writing. Covers CONSORT/STROBE/PRISMA and report structure. For evidence quality see scientific-critical-thinking; scoring see scholar-evaluation." license: CC-BY-4.0
Peer review is a systematic process for evaluating scientific manuscripts and grant proposals. This knowhow covers the complete review cycle: from initial assessment through detailed section-by-section evaluation, statistical rigor checks, reproducibility assessment, figure integrity verification, ethical considerations, and writing quality — culminating in a structured review report with actionable feedback.
| Review Type | Scope | Typical Length | Key Focus | |-------------|-------|---------------|-----------| | **Original research** | Full evaluation | 1–3 pages | Rigor, novelty, reproducibility | | **Review/Meta-analysis** | Coverage and bias | 1–2 pages | Completeness, search strategy, systematic approach | | **Methods paper** | Validation | 1–2 pages | Comparison to existing methods, reproducibility | | **Short communication** | Proportional | 0.5–1 page | Core findings rigor despite brevity | | **Grant proposal** | Feasibility + significance | 1–3 pages | Innovation, approach, team, budget justification |
| Standard | Applies to | Key Check | |----------|-----------|-----------| | **CONSORT** | Randomized controlled trials | Flow diagram, randomization, blinding, ITT analysis | | **STROBE** | Observational studies | Study design, setting, participants, variables, bias | | **PRISMA** | Systematic reviews / meta-analyses | Search strategy, PICO, risk of bias, forest plots | | **ARRIVE** | Animal research | Species, sample size, randomization, 3Rs | | **MIAME** | Microarray experiments | Platform, normalization, data deposit (GEO/ArrayExpress) | | **MINSEQE** | Sequencing experiments | Read counts, mapping, QC metrics, data deposit (SRA) |
What are you reviewing?
├── Scientific manuscript
│ ├── Original research → Full 7-stage workflow
│ ├── Review / meta-analysis → Emphasize Stage 2 (Introduction + Methods) + Stage 4
│ ├── Methods paper → Emphasize Stage 3 (rigor) + Stage 4 (reproducibility)
│ └── Short communication → Abbreviated workflow (Stages 1, 2, 3, 7)
├── Grant proposal
│ └── Focus on: significance, innovation, approach feasibility, team qualifications
└── Not a document review
└── Use scientific-critical-thinking for evidence evaluation| Reviewer Situation | Focus Areas | Time Budget | |-------------------|-------------|-------------| | First-round journal review | All 7 stages, full detail | 4–8 hours | | Revision re-review | Only check if previous concerns addressed | 1–2 hours | | Internal lab feedback | Stages 1–3, 7 (skip ethics, reporting standards) | 2–4 hours | | Conference abstract | Stage 1 only + brief methods check | 30 min | | Grant review | Significance + innovation + approach + team | 3–6 hours |
1. **Read the whole manuscript once before writing any comments**: Resist the urge to annotate during first reading. Your initial impression informs the summary statement and helps distinguish major from minor issues.
2. **Separate description from judgment**: For each concern, first state what you observed ("The authors report p=0.04 without correction for 12 comparisons"), then state why it's problematic ("This inflates the false positive rate"), then suggest a fix ("Apply Bonferroni or FDR correction and re-evaluate significance").
3. **Number every comment for easy reference**: Both major and minor comments should be sequentially numbered so authors can address each one explicitly in their response letter.
4. **State your confidence level on statistical concerns**: If you are uncertain about a statistical issue, say so explicitly ("I am not certain whether the normality assumption holds here — the authors should verify with a Shapiro-Wilk test or use a non-parametric alternative").
5. **Check reporting standards compliance**: Before writing the review, identify which reporting standard applies (CONSORT, STROBE, PRISMA, etc.) and use its checklist to verify completeness. Missing checklist items are legitimate major comments.
6. **Acknowledge strengths explicitly**: Every review should include 2–3 specific strengths. This is not politeness — it tells editors which aspects are sound and helps authors understand what NOT to change during revision.
7. **Never request experiments beyond the study's scope**: A reviewer should point out limitations, not redesign the study. "This limitation should be acknowledged in the Discussion" is appropriate. "The authors should run a new cohort with 500 patients" is not.
8. **Be concrete about figure quality**: Instead of "figures need improvement," say "Figure 3A: y-axis label missing units; error bars are not defined in the legend (SD vs SEM?); color coding is not colorblind-accessible."
9. **Maintain anonymity and professionalism**: In single/double-blind review, avoid self-referential comments ("In our previous work..."). Never use dismissive or condescending language. Frame all criticism constructively.
10
Turn your AI coding agent into a life sciences expert — 199 bioinformatics skills for Claude Code covering RNA-seq, single-cell analysis, genomics, proteomics, drug discovery, and more. Boosted BixBench from 65% to 92%. Open source.
Scaffold a new SciAgent-Skills entry. Picks pipeline/toolkit/database/guide template, creates skills/{category}/{name}/SKILL.md with valid frontmatter, appends…
Bayesian modeling with PyMC 5: priors, likelihood, NUTS/ADVI sampling, diagnostics (R-hat, ESS), LOO/WAIC comparison, prediction. Hierarchical, logistic, GP…
Time-to-event modeling with scikit-survival: Cox PH (elastic net), Random Survival Forests, Boosting, SVMs for censored data. C-index, Brier, time-dependent…
Guided statistical analysis: test choice, assumption checks, effect sizes, power, APA reporting. Pick tests, verify assumptions, or format results for…
Python statistical modeling: regression (OLS, WLS, GLM), discrete (Logit, Poisson, NegBin), time series (ARIMA, SARIMAX, VAR), with rigorous inference,…
DL cell/nucleus segmentation for fluorescence and brightfield microscopy. Pre-trained models (cyto3, nuclei, tissuenet) and a generalist flow-based algorithm…