| name | scientific-paper |
| description | End-to-end workflow for creating, reviewing, and critiquing scientific papers with rigorous methodology. Use when: writing papers, reviewing papers, literature reviews. Do not use: for blog posts, technical documentation. |
Scientific Paper Skill
Complete workflow for empirical SE research: paper creation, deep review, SLR, methodology critique, peer-review simulation, and LaTeX conventions.
Modes of Operation
| Mode | Trigger phrases | What it does |
|---|
create | "create paper", "new paper", "paper template", "criar artigo" | Scaffolds paper structure + LaTeX template |
review | "review paper", "revisar paper", "find errors" | Full structured review: methodology, writing, validity |
critique | "critique", "metodologia", "methodological errors", "erros metodológicos" | Deep methodology critique using error catalog |
peer-review | "peer review", "referee report", "simulate reviewer", "parecerista" | Generates structured referee report (Summary / Strengths / Weaknesses / Questions / Recommendation) |
slr | "systematic review", "SLR", "revisão sistemática", "literature review" | Guides SLR protocol from scratch |
methodology-check | "check methodology", "rigor", "validity", "research design" | Runs methodology checklist for study type |
data-presentation | "figures", "tables", "charts", "data viz", "plots", "benchmark results" | Reviews data presentation choices |
bib-validate | "validate bibliography", "check citations", "missing citations", "bib" | Runs validate-bib.py to catch orphaned \cite{} and missing .bib entries |
Step-by-Step Procedures
Mode: review - Full Paper Review
- Ask for the paper file or content (paper.tex or text paste).
- Load error catalog to identify issues by category.
- Load methodology checklist and run the appropriate checklist for the paper's study type.
- Load threats-to-validity to audit the validity section.
- Load paper-structure to verify completeness of all sections.
- Produce structured output:
- Critical errors (must fix before submission)
- Major concerns (likely desk-reject or major revision)
- Minor issues (polish, clarity, consistency)
- Positive observations (what is done well - important for balance)
- For each issue: location (section/paragraph) → problem → recommended fix.
Mode: peer-review - Referee Report Simulation
- Ask for paper content + target venue (EMSE, ICSE, FSE, MSR, arXiv, etc.).
- Load venues.md for venue-specific criteria and expectations.
- Load error-catalog.md and threats-to-validity.md.
- Write a complete referee report in the style of the target venue:
SUMMARY
[2–3 sentences: what the paper does and its main claim]
STRENGTHS
1. ...
2. ...
3. ...
WEAKNESSES / MAJOR CONCERNS
1. ...
2. ...
MINOR COMMENTS
1. ...
SPECIFIC QUESTIONS FOR AUTHORS
1. ...
RECOMMENDATION
[ ] Accept [ ] Minor Revision [X] Major Revision [ ] Reject
CONFIDENCE: [Reviewer expertise 1–5]
- Provide a separate "author's perspective" note: what changes would move the paper from current recommendation to Accept.
Mode: critique - Methodology Critique
- Identify the study type from the paper (case study / benchmarking / experiment / SLR / survey).
- If the paper is a case study, use the
scientific-case-study-research skill for case definition, units of analysis, protocol, triangulation, chain of evidence, ethics, and threats to validity.
- Load methodology-checklist.md for that type.
- Run each checklist item and flag gaps.
- Load error-catalog.md and map identified issues to error categories.
- Prioritize by severity: conclusion validity > internal validity > construct validity > external validity.
Mode: slr - Systematic Literature Review
- Load systematic-review-protocol.md.
- Walk through each phase: motivation → RQs → search string → databases → inclusion/exclusion → quality assessment → data extraction → synthesis.
- Use assets/slr-protocol-template.md to scaffold the protocol document.
- Validate search string completeness (synonyms, acronyms, Boolean operators).
- Check that each RQ maps to a data extraction field and a synthesis method.
Mode: create - New Paper Scaffold
- Determine study type and target venue.
- Load paper-structure.md for the appropriate template.
- Use assets/paper-template.tex as the LaTeX starting point.
- Load latex-conventions.md for repo-specific setup.
- Load venues.md for venue formatting requirements.
- Scaffold folder:
papers/YYYY-short-slug/ with paper.tex, title-page.tex, README.md.
Mode: data-presentation - Figures and Tables Review
- Load data-presentation.md.
- Audit each figure/table: chart type appropriateness, axis labels, units, error bars, statistical significance markers, color accessibility, caption completeness.
- Flag misleading visualizations (e.g., truncated y-axis, missing baseline, no variance shown).
Mode: bib-validate - Bibliography Validation
- Locate
paper.tex and *.bib file in the workspace.
- Run validate-bib.py:
python scripts/validate-bib.py --tex paper.tex --bib references.bib
- Report:
- Orphaned
\cite{}: keys used in .tex but missing from .bib
- Unused entries: keys in .bib never cited in .tex
- Malformed entries: missing required BibTeX fields (author, title, year)
- Run after any round of work where new citations were added; skip if no new
\cite{} keys were introduced.
Key Resources
Write like a human
Even in formal academic prose, keep out the AI writing tells reviewers now flag:
- No em dashes and no curly quotes; use straight quotes.
- Cut AI vocabulary: "delve", "leverage", "robust", "crucial", "pivotal", "testament", "underscore", "showcase", "intricate", "meticulous", "landscape" (as an abstract noun).
- Remove significance padding ("stands as a testament to", "plays a pivotal role"), trailing "-ing" analysis clauses ("...highlighting its significance"), and copula avoidance ("serves as" for "is").
- Do not use "novel" or "state-of-the-art" without a named comparator. Attribute claims to a citation, not to "experts" or "studies".
- Open sections with the finding or claim, not "In this paper" or "It is important to note". Do not close sections with "In summary"/"In conclusion" restatements.
