Paper Writing
Turn research notes into a structured academic paper with abstract, contributions, and section-level guidance your agent can draft section by section.
Overview
Paper Writing is an agent skill most often used in Build (also Idea, Validate) that structures academic papers with abstract, introduction, and explicit contribution bullets for agent-assisted drafting.
Install
npx skills add https://github.com/vishalsachdev/claude-code-skills --skill paper-writingWhat is this skill?
- Sectioned template from title block through abstract, introduction, and background
- Contribution paragraph pattern with four numbered contribution slots
- Abstract scaffold targeting 150–300 self-contained words
- Introduction flow: hook, context, gap, key insight, then explicit contributions
- Inline editorial comments guide pacing (e.g. get to contribution in 1–1.5 pages)
- Abstract guidance targets 150–300 words
- Introduction pacing note: reach contributions within about 1–1.5 pages
- Four templated contribution slots in the contributions paragraph
Adoption & trust: 957 installs on skills.sh; 4 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
You have research or engineering results but no consistent manuscript structure, so agent drafts wander and miss contributions, bounds, and reproducibility details reviewers expect.
Who is it for?
Solo researchers, indie technical founders, and students writing papers, theses, or formal reports with an AI coding agent as co-author.
Skip if: Quick marketing landing copy, API reference-only docs, or cases where the spec is already a finalized LaTeX template with no structural help needed.
When should I use this skill?
You are drafting or revising an academic or formal technical paper and need section-by-section structure, contribution framing, and abstract constraints.
What do I get? / Deliverables
You get a section-ordered paper template with placeholder guidance filled into publishable prose, including a clear contributions list and abstract within stated word bounds.
- Structured manuscript draft with abstract and introduction
- Explicit numbered contributions list tied to sections and artifacts
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
The canonical shelf is build/docs because the primary deliverable is a finished manuscript artifact, even though planning happens earlier in the journey. Subphase docs fits long-form technical writing with fixed section templates, contribution bullets, and reproducibility blocks—not frontend code or API implementation.
Where it fits
Outline contribution bullets and related work gaps before committing to a benchmark sprint.
Turn prototype metrics into an abstract and results placeholders to see if the story is publication-viable.
Draft full introduction and background sections from repo notes while the agent keeps section comments and citation slots consistent.
How it compares
Use as a procedural writing template instead of unstructured 'write my paper' prompts that omit contribution and evaluation sections.
Common Questions / FAQ
Who is paper-writing for?
Solo and indie builders who produce academic or formal technical papers and want their agent to follow a standard IMRaD-style scaffold with explicit contributions.
When should I use paper-writing?
In idea/research when outlining a submission, in validate when turning a prototype evaluation into a write-up, and in build/docs when drafting or revising full manuscript sections from notes.
Is paper-writing safe to install?
It is a local writing template skill without bundled network calls; still review the Security Audits panel on this Prism page and avoid pasting unpublished secrets into drafts.
SKILL.md
READMESKILL.md - Paper Writing
# [Title: Make It Clear and Specific] **Authors:** Author Name¹, Author Name², Author Name¹ **Affiliations:** ¹Institution 1, ²Institution 2 **Contact:** corresponding.author@email.com --- ## Abstract <!-- 150-300 words, self-contained summary --> [**Background (1-2 sentences):**] [Why this problem matters] [**Problem (1 sentence):**] [Specific problem addressed] [**Solution (1-2 sentences):**] [Your key insight and approach] [**Results (2-3 sentences):**] [Main findings with specific numbers] [**Impact (1 sentence):**] [Significance and implications] --- ## 1. Introduction <!-- Get to your contribution quickly (1-1.5 pages) --> [**Hook paragraph:**] [Why should anyone care about this problem?] [**Context paragraphs:**] [What's the broader landscape? What have others done?] [**Gap paragraph:**] [What's missing? Why existing solutions fall short?] [**Key insight:**] [What's your core idea that makes this work?] [**Contributions paragraph:**] This paper makes the following contributions: • **[Contribution 1]:** [Novel technique/algorithm/framework] that [what it achieves] by [key mechanism] • **[Contribution 2]:** [Theoretical analysis/proof] showing [property] with [bound/complexity] (Section X, Theorem Y) • **[Contribution 3]:** [Empirical evaluation] demonstrating [performance gain] on [benchmarks] compared to [baselines] (Section X) • **[Contribution 4]:** [Open-source release] of [code/data/model] at [URL] for reproducibility --- ## 2. Background and Problem Formulation <!-- Define the problem formally (0.5-1 page) --> ### 2.1 Problem Definition [Formal problem statement] **Given:** [Input description] **Goal:** [Objective, possibly with formal notation] **Constraints:** [Any constraints or requirements] ### 2.2 Notation and Preliminaries **Table 1:** Notation | Symbol | Meaning | |--------|---------| | $n$ | [Description] | | $G = (V, E)$ | [Description] | | $\mathcal{D}$ | [Description] | [Any background concepts needed to understand your approach] --- ## 3. Approach <!-- Core technical content (2-3 pages) --> ### 3.1 Overview [High-level intuition in plain language] **Figure 1:** System architecture / Overview diagram ``` [Diagram showing overall approach] ``` ### 3.2 [Component/Algorithm 1] [Detailed description of first major component] **Algorithm 1:** [Algorithm Name] ``` Input: [inputs] Output: [outputs] 1: procedure NAME(parameters) 2: initialize [variables] 3: while [condition] do 4: [step 1] 5: [step 2] 6: end while 7: return [result] ``` [Explanation of algorithm logic and why it works] ### 3.3 [Component/Algorithm 2] [Detailed description of second major component] ### 3.4 [Integration/Full System] [How components work together] ### 3.5 Theoretical Analysis (if applicable) **Theorem 1** (Complexity): [Statement] *Proof sketch:* [Key ideas of the proof, or defer to appendix] **Theorem 2** (Correctness): [Statement] *Proof:* [Proof or reference to appendix] --- ## 4. Experimental Evaluation <!-- Demonstrate that it works (2-3 pages) --> ### 4.1 Experimental Setup **Datasets:** We evaluate on [N] benchmark datasets: - **Dataset A** [citation]: [Size, characteristics, why it's relevant] - **Dataset B** [citation]: [Size, characteristics, why it's relevant] - **Dataset C** [citation]: [Size, characteristics, why it's relevant] **Baselines:** We compare against [N] state-of-the-art methods: - **Method 1** [citation]: [Brief description, year] - **Method 2** [citation]: [Brief description, year] - **Method 3** [citation]: [Brief description, year] **Metrics:** - **[Metric 1]:** [Why appropriate for this problem] - **[Metric 2]:** [Why appropriate for this problem] - **[Metric 3]:** [Runtime / scalability] **Implementation Details:** - Platform: [Python 3.10, PyTorch 2.0, etc.] - Hardware: [GPU model, CPU specs] - Hyperparameters: [Learning rate α = X, batch size = Y, etc.] - Code: Available at [URL] ### 4.2 Main Results *