/ieee-writing

name: ieee-writing
description: >-
  Draft, restructure, or plan IEEE communications-journal manuscript sections (JSAC, TWC,
  TC/TCOM, WCL, CL): abstract + Index Terms, introduction with a contributions list, related
  work, system model, problem formulation, proposed solution/algorithm, performance analysis
  (theorems/proofs), numerical results, discussion, conclusion, title — from author claims,
  derivations, simulation results, figures, notes, or Chinese drafts. Use this whenever the user
  wants to WRITE or REBUILD comms paper text: "help me write my paper", "draft the introduction",
  "写系统模型", "问题建模", "restructure this section", "把我的中文草稿写成英文论文", "写贡献点", "改写算法部分" — not merely fix
  grammar (use ieee-polishing). For a WCL/CL letter, use ieee-letter. Bilingual: turns Chinese
  notes into submission-ready English plus short Chinese explanations.

IEEE-Style Scientific Writing

Use this skill to **create or rebuild** IEEE manuscript prose — to construct the argument, not only to polish finished sentences. For sentence-level language work, hand off to `ieee-polishing`.

Core stance

• **Author evidence comes first.** Never invent results, numbers, mechanisms, references,

benchmark schemes, channel models, simulation parameters, network architectures, training data, novelty claims or limitations. If essential evidence is missing, write an explicit `[PLACEHOLDER: ...]` or ask the user — do not fill the gap.

• **Write the argument before the sentences.** Decide what each paragraph must prove, then draft.
• **Make the paper easy to judge** on relevance, novelty, soundness, and reproducibility —

these are the axes IEEE reviewers actually score.

• **Ambitious but bounded claims.** Tie every claim to the evidence that supports it and stop

the claim where the evidence stops.

• **IEEE register**, not magazine narrative: precise, technical, US English, numbered citations

`[n]`, equations numbered only for the skeleton of the argument.

When to open extra files

Keep this file lean. Open a reference only when its row matches the task.

| File | Open when | |---|---| | [references/journal-fit.md](references/journal-fit.md) | The user names or is choosing JSAC, TWC, TC/TCOM, WCL, or CL; deciding whether a draft fits a transaction vs. a letter; adapting argument depth, evidence depth, or page strategy to the target journal | | [references/abstract.md](references/abstract.md) | Drafting/revising the abstract and Index Terms | | [references/introduction.md](references/introduction.md) | Drafting/revising the Introduction, the technical-challenge build-up, the contributions list, or the paper-organization paragraph | | [references/method.md](references/method.md) | Writing the System Model, Problem Formulation, and Proposed Solution/Algorithm: signal flow, optimization problem, transformations, notation, equation-numbering, convergence/complexity | | [references/learning-based-comms.md](references/learning-based-comms.md) | The contribution is a neural network or learning algorithm for a PHY/MAC task (DL/CNN/GNN/Transformer, deep unfolding, DRL): architecture, training methodology, loss tied to a comms metric, generalization and inference-complexity claims | | [references/system-model-paradigms.md](references/system-model-paradigms.md) | Writing the System Model for an emerging architecture (RIS, NOMA, ISAC, near-field/XL-MIMO, movable/fluid antenna, cell-free, mmWave/hybrid, semantic comm): the must-have model elements, the distinctive variable/constraint, the signature metric, and per-paradigm pitfalls | | [references/theory-and-proofs.md](references/theory-and-proofs.md) | Writing the optimization solution (SCA/SDR/ADMM/fractional programming) or the Performance Analysis: theorem/lemma/proof structure, appendix placement, convergence and complexity claims | | [references/results-discussion-conclusion.md](references/results-discussion-conclusion.md) | Writing the Numerical Results narrative, Discussion interpretation, or a bounded Conclusion | | [references/body-revision.md](references/body-revision.md) | Tightening, compressing, or restructuring existing body text; fixing logic flow, term consistency, table/prose overlap, or task-hierarchy confusion | | [references/chinese-author-workflow.md](references/chinese-author-workflow.md) | The user's input is Chinese, mixed Chinese-English, or lab-note style rather than manuscript prose |

For simulation *content* (which benchmark schemes/metrics/sweeps to run) use the `ieee-experiments` skill; for figures use `ieee-figure`; for reference formatting use `ieee-citation`; for a WCL/CL letter use `ieee-letter`.

Intake — establish before drafting

• **section**: title, abstract, Index Terms, introduction, related work, system model, problem

formulation, proposed solution/algorithm, performance analysis, numerical results, discussion, conclusion, or full outline

• **paper type**: optimization/resource-allocation design, performance analysis (closed-form

outage/BER/rate), new system/architecture (RIS, NOMA, ISAC, cell-free, ...), learning-based (DL/deep-unfolding/DRL for a PHY/MAC task), signal-processing algorithm, or measurement/empirical study

• **core claim**: what the paper actually demonstrates, in one sentence
• **evidence**: derivations/theorems, simulation curves, benchmark comparisons, complexity analysis
• **boundary**: where the claim stops (CSI assumption, channel model, scope, conditions)
• **target**: the specific journal (JSAC/TWC/TCOM = transactions; WCL/CL = letter) and its current

page/abstract rules; if the target is unclear, use `references/journal-fit.md` before drafting

If `target`, `core claim`, `evidence`, or `boundary` is missing, surface the gap first. You may still produce a scaffold with explicit placeholders.

Writing workflow

1. **One-sentence argument:** `In [problem/system], we show [advance] using [approach], supported by [evidence], under [boundary].` Get the user to confirm i