CommunityResearch & Data Analysisgithub.com

informs-writing

Write, structure, rewrite, and polish academic manuscripts in LaTeX (informs4) for INFORMS flagship journals: Operations Research (OR), Management Science (MS), Manufacturing & Service Operations Management (M&SOM), and INFORMS Journal on Computing (IJOC). Use when drafting a paper or section from a core idea via exemplar papers (arXiv/SSRN/top-journal PDFs), rewriting drafts to remove AI prose patterns, pruning excessive notation and redundant definitions, fixing context-blind concepts and revision-history leakage, formatting with informs4.cls and Online Appendix conventions, or calibrating manuscripts to OR, MS, M&SOM, or IJOC editorial preferences.

What is informs-writing?

informs-writing is a Claude Code agent skill that write, structure, rewrite, and polish academic manuscripts in LaTeX (informs4) for INFORMS flagship journals: Operations Research (OR), Management Science (MS), Manufacturing & Service Operations Management (M&SOM), and INFORMS Journal on Computing (IJOC). Use when drafting a paper or section from a core idea via exemplar papers (arXiv/SSRN/top-journal PDFs), rewriting drafts to remove AI prose patterns, pruning excessive notation and redundant definitions, fixing context-blind concepts and revision-history leakage, formatting with informs4.cls and Online Appendix conventions, or calibrating manuscripts to OR, MS, M&SOM, or IJOC editorial preferences.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/JiayuXu11/informs-writing-skills/tree/HEAD/informs-writing

Installed? Explore more Research & Data Analysis skills: affaan-m/uspto-database, affaan-m/scholar-evaluation, affaan-m/literature-review · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

INFORMS Flagship Journal Writing & Editing (informs-writing)

Write, structure, rewrite, and referee-harden academic manuscripts for Operations Research (opre), Management Science (mnsc), Manufacturing & Service Operations Management (msom), and INFORMS Journal on Computing (ijoc) in unified informs4 LaTeX.


Core Operational Principles

  1. Git Version Tracking First: Before substantive drafting or rewriting, confirm the workspace is Git-initialized with .gitignore from latex-gitignore.txt (check with git status; initialize if needed). Commit clean checkpoints per section or revision pass (e.g., draft(intro): outline and first pass, polish(model): prune unused notation) so every prose and equation edit is auditable and reversible.
  2. Source-Fidelity & Protected-Regions Contract: Preserve every factual claim, theorem statement, parameter symbol, assumption label, table number, \cite{...}, \ref{...}, \eqref{...}, and .tex macro. Never invent citations, proofs, robustness checks, or fake "humanizer" tics. Treat code, LaTeX math ($...$, \[...\], equation, align), booktabs tables, and verbatim quotes as protected regions unless doing an explicit mathematical notation audit.
  3. Silent 4-Pass Discipline (<= 2 Iterative Passes): Diagnose internally and execute cleanly without narrating process steps inside manuscript prose. Stop after at most two focused revision passes per section unless the user requests an additional pass; never over-scrub domain-standard mathematical prose into conversational fluff.
  4. Pure Reader-Facing English: Target manuscript prose has zero memory of chat prompts, LLMs, or earlier drafts. Never leak meta-commentary ("Unlike the previous version...", "As requested...") or template placeholders (TODO, FIXME, [INSERT ...]) into .tex files.
  5. Visual Aesthetics & Compiled-PDF QA: Never judge typography, tables, or figures from .tex source alone. Build crisp booktabs tables (@{}...@{}, \arraystretch, \cmidrule(lr)) and vector figures sized for 8–10 pt labels, compile the PDF whenever a TeX engine is available, run scripts/check_pdf_layout.py, and visually inspect flagged pages via read_file to catch blank voids, overlaps, margin overflows, or garbled glyphs.

Modes & Reference Routing

Select the primary mode matching the user's request, load only the relevant reference modules below, and follow their procedures step-by-step:

ModeWhen to UseLoad These References First
Mode A: 0 -> 1 Drafting via ExemplarsStarting from a core idea, problem, or outline; retrieving legal arXiv/SSRN/faculty-page PDFs to mirror section pacing, sentence-by-sentence Abstract architecture (< 270 words), <= 3-page Introduction budget, corresponding section word-count parity (+/-15%--20% vs. exemplars), theorem-to-prose ratios, and paragraph cadence.workflow-write-exemplar.md, section-and-lane-architecture.md, journals-informs-matrix.md
Mode B: Anti-AI Rewriting & Prose SurgeryPolishing an AI-generated or stiff human draft; pruning notation bloat; fixing cold-start concepts, defensive caveats, raw decimal dumping, definition drift, or chat leakage.workflow-rewrite-deai.md, human-craft-and-anti-ai.md, math-notation-and-appendix.md
Mode C: Referee Calibration, Contract Audit & Journal RetargetingChecking cross-section claim consistency, calibrating evidence verbs, preparing identification specs, converting a working paper for submission, or answering referees.manuscript-contract-and-reviewer-calibration.md, empirical-and-identification-craft.md, journals-informs-matrix.md

Supporting resources (cross-mode):


Seven Hard-Blocked Iron Rules

scripts/audit_informs_draft.py enforces these rules as hard-blocking ERROR gates (exit code 1):

  1. Disease 1 — Notation Minimalism & Equation Hygiene: Define only symbols reused across multiple expressions or carrying substantive economic/operational meaning. Every display equation, subscript, and parameter must earn its keep. Number only equations referenced via \eqref/\ref; use align*/\[...\] otherwise; forbid eqnarray.
  2. Disease 2 — Context-First Introduction: Never cold-start a technical term, stylized setting, \begin{definition}, \begin{theorem}, or notation before intuitive motivation and operational meaning in plain English.
  3. Disease 3 — Natural, Cadenced Scholarly English: Eliminate AI vocabulary clichés (delve, tapestry, underscore, pivotal, moreover/furthermore crutches), participial tail clauses (..., thereby enabling...), Chinglish calques (with the rapid development of...), prose em-dashes (--- / —), decorative colon-label bullets, and negative parallelisms (not merely X, but Y). Use concrete actor-verb-object sentences.
  4. Disease 4 — Result Confidence (No Defensive Caveats): State structural limits cleanly once inside Scope / Model Setup or Future Work. Never undermine a theorem, managerial insight, or empirical finding right after presenting it ("Although we prove X, this does not fully establish...").
  5. Disease 5 — Pattern Prose vs. Table Precision: Narrate economic mechanisms, orders of magnitude, clean rounded benchmarks (roughly 1%--2%, nearly threefold), and qualitative comparative statics in prose; keep exact 2--3 decimal point estimates, standard errors, and confidence intervals inside booktabs tables (using threeparttable + siunitx, zero \hline, zero vertical |).
  6. Disease 6 — Strict Terminology Discipline: Define each core construct exactly once; use one canonical term and symbol consistently throughout the paper (no synonym cycling across Abstract, Introduction, Model, Results, and Conclusion).
  7. Disease 7 — Zero Chat / Revision-History Leakage: Never write "Unlike the previous draft", "I have corrected", "per reviewer feedback", or unfilled bracket/TODO placeholders inside any manuscript file. Keep referee response letters in separate standalone documents.

Two-Layer Journal Baseline & Deltas (informs4)

Full specifications live in journals-informs-matrix.md. Summary baseline + binding journal deltas:

  • Shared Baseline (All 4 Journals): Unified informs4.cls (nonblindrev for review, blindrev / double-anonymous where required); 100% Unified Online Appendix (\ECSwitch) with zero regular \appendix in the main body; booktabs + threeparttable tables (\toprule, \midrule, \bottomrule); zero prose em-dashes (en-dash -- reserved strictly for numeric/page ranges); calibrated evidence verbs (characterize, establish, bound, document, identify, estimate, is consistent with).
OptionJournalAbstract & KeywordsMain-Body Length NormAppendix & Proof PolicyEditorial Signature & Hard Rules
opreOperations ResearchSingle-block \ABSTRACT + \SUBJECTCLASS + \AREAOFREVIEW~32–36 pp. single-spaced (<= 35 pp. target incl. refs)100% \ECSwitch (Online Appendix EC.x); pair headline theorems with 3–6 sentence body Proof SketchRigor + surprising methodological or structural payoff; no orphaned formal results
mnscManagement ScienceSingle-block \ABSTRACT (< 270 words; check portal limit) + 3–5 lowercase \KEYWORDS + \HISTORY<= 32 pp. (1.5-spaced informs4 single-col incl. refs)100% \ECSwitchBroad cross-functional significance; Data & Code Disclosure section; clear practical/managerial insight
msomM&SOMMandatory 3-part structured \ABSTRACT (Problem definition, Methodology/results, Managerial implications)<= 32 pp. manuscript100% \ECSwitch (16-page target ceiling for EC)Endnotes (\theendnotes) instead of bottom-of-page \footnote; tight end-to-end operations grounding
ijocINFORMS J. on ComputingConcise single-block \ABSTRACT + problem/computational impact~30–35 pp. incl. refs100% \ECSwitch (Online Appendix EC.x); algorithmic complexity + key proof sketch in bodyOR + CS fusion; formal algorithm2e pseudocode; .cd Dual-DOI Software & Data companion repo protocol (INFORMSJoC)

Automated Quality Gate (Run Before Every Final Handoff)

Run all three verification scripts from the skill root (or pass full paths to your manuscript .tex / text):

  1. Hard LaTeX & AI-Disease Gate (exits 1 on any hard ERROR; pass --advisory for non-blocking inspection):
    python scripts/audit_informs_draft.py path/to/main.tex --journal mnsc
    
  2. Paragraph Craft, Verb Calibration & Rhythm Check (supports file path or stdin, --section, --strict):
    python scripts/check_prose_craft.py path/to/section.tex --section intro --strict
    
  3. Cross-Section Contract Drift, Exemplar Length Parity & Online Appendix Placement (.tex or .md/.txt):
    python scripts/audit_manuscript_contract.py path/to/main.tex --journal opre --exemplar-abstract-words <N_abs> --exemplar-intro-words <N_intro>
    python scripts/audit_manuscript_contract.py --place-results items.txt --journal ijoc
    
  4. Compiled-PDF Visual Layout, Whitespace & Glyph Gate (always inspect rendered PNGs via read_file):
    python scripts/check_pdf_layout.py path/to/main.pdf --render-flagged /tmp/pdf_qa
    

After all source and PDF gates pass cleanly, stage changed .tex files and create a descriptive Git commit before delivering the updated manuscript.

Related Skills