Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Paper Writing: Evidence-Grounded Academic Revision

    1

    Draft and revise research papers with clear arguments, precise prose, and consistent evidence.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Paper Writing: Evidence-Grounded Academic Revision

    Paper Writing: Evidence-Grounded Academic Revision

    Example session with this skill installed

    Revise the abstract below into one coherent paragraph of 100–130 words. Keep the supplied numbers and evaluation scope. Then give a short explanation of substantive edits and any claims that need more evidence.

    This is a fictional worked example. Its results are illustrative, not findings from a real paper.

    Verified facts within this example

    • A benchmark of 1,200 paired images for object localization under partial visibility.
    • Object visibility annotations and held-out evaluation splits.
    • A reference model trained with additional views available only during training.
    • Accuracy on the 300-example held-out split rises from 62.0% to 68.5% under matched evaluation.
    • No component ablation or cross-dataset testing was run.

    Draft abstract
    Object localization with partial visibility is very important. We propose a revolutionary unified framework called ViewBench. It includes 1,200 paired images, object visibility annotations, held-out splits, and a teacher-student model with a training-only additional-view branch. Our benchmark solves occlusion and provides robust generalization across all scenes. The accuracy increases from 62.0% to 68.5% on 300 held-out examples. This proves that the additional-view branch is responsible for the entire gain. The model and benchmark are excellent and useful for future research.

    • Read your context and instructions
    • Compiled the paper writing

    Revised abstract

    In this fictional worked example, ViewBench is a benchmark for object localization under partial visibility, comprising 1,200 paired images, object visibility annotations, and held-out evaluation splits. The paired images support training with additional views, while the annotations and splits define the benchmark’s evaluation setting. To illustrate its use, we evaluate a reference model with additional views available only during training. Under matched evaluation on the 300-example held-out split, accuracy rises from 62.0% to 68.5%. These illustrative results provide a reference comparison for the complete model configuration and demonstrate how the benchmark can be used to evaluate localization under partial visibility in a defined setting.

    Editorial notes

    The revision centers the benchmark, connects its contents to its use, and preserves the numbers, training-only additional views, and matched evaluation scope. It replaces promotional language and universal claims with the supported comparison. Attributing the entire gain to the additional-view branch would require a controlled component ablation; claims of cross-dataset generalization would require testing on other datasets. Neither was performed. The results remain explicitly fictional and illustrative.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Turn a draft into a coherent research argument

    Paper Writing helps your AI agent draft, restructure, and refine research papers while keeping the claims tied to the evidence you provide. It supports focused paragraph edits, section revisions, and consistency checks across a manuscript.

    It is designed for researchers, graduate students, and technical authors who need to explain what their work contributes, how it is evaluated, and what the results establish.

    What it helps you do

    • Rewrite abstracts and introductions around the research question, contribution, and supported findings.
    • Explain methods, datasets, and benchmarks without forcing every paper into the same template.
    • Improve related-work structure, technical prose, transitions, and terminology.
    • Write experiment discussions and result captions with the correct metrics and comparison scope.
    • Check whether claims, symbols, numbers, training conditions, and test-time inputs agree across sections.
    • Separate manuscript-ready text from editorial notes about missing evidence.

    What you receive

    A complete SKILL.md package with Codex interface metadata and eight focused reference guides: Abstract and Introduction; Benchmark and Dataset Exposition; Related Work; Method and Technical Exposition; Experiments and Result Captions; Conclusion and Limitations; Local Prose; Whole-Manuscript Consistency.

    How to use it

    Install the skill, provide your draft and relevant evidence, and specify the edit you want. For example:

    "Use paper-writing to revise this Introduction so the problem, contribution, and evaluation connect clearly. Preserve all verified results and references."

    The agent returns revised text and a concise explanation of substantive edits and unresolved evidence gaps. File edits and rendering checks depend on the tools available in your agent.

    Ongoing maintenance

    This marketplace edition will be maintained with refinements to the writing workflow, reference guides, and compatibility as needs evolve. Future updates will be published here first. The public GitHub version will remain available and may not include every subsequent marketplace update.

    Updates released for this listing are available to existing buyers through Agensi.

    Open-source availability

    Created and maintained in codex-skills-library. The initial release matches the core skill currently available free on GitHub under the MIT license:

    https://github.com/sidiangongyuan/codex-skills-library/tree/main/skills/paper-writing

    This listing provides a ready-to-install package with ongoing marketplace maintenance. The package retains its MIT license.

    Scope and limitations

    You supply the scientific content and evidence. Output quality depends on that material and the host model. Citation verification requires access to primary sources; this workflow does not guarantee error-free references or publication acceptance. Peer-review risk assessment, official rebuttal management, and new experiment design are separate tasks.

    Needs accurate source material and relevant evidence. Cannot establish untested scientific claims or guarantee publication acceptance. Citation checks, compilation, and visual QA depend on the tools available in the host agent. Optional companion skills mentioned in the instructions are not included and are not required for core writing tasks.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Permissions

    Write Files
    Read Files
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill acts as a specialized editor that restructures your draft around its core findings, ensuring that technical prose, experimental results, and method descriptions are logically consistent and evidence-grounded.

    You receive a complete SKILL.md package with metadata and eight specialized reference guides covering every section of a manuscript, from the abstract to the limitations and results captions.

    Yes, the skill is designed to handle specific technical requirements including dataset benchmarks, mathematical notation, and result consistency across different sections of a paper.

    This skill is compatible with any AI agent that supports the Codex interface; simply install the package and provide your draft along with the relevant evidence or data.

    No, the skill relies on the evidence and data you provide; while it helps structure and refine citations, it does not browse the web to find new primary sources for you.

    Marketplace buyers receive exclusive ongoing maintenance and compatibility updates through the platform, which may include refinements not found in the initial open-source release.

    New in Writing & Content