Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Paper Writing Agent

    1

    The review loop brings three perspectives: methods, evidence, and significance.

    Free

    Security scanned
    Paper Writing Agent

    Paper Writing Agent

    Example session with this skill installed

    Synthetic test example: Open my existing LaTeX draft about a toy cache. It has main.tex and main.pdf, with no measured results. Propose a review plan using methods, evidence, and significance perspectives. Explain what each job will receive, keep the manuscript unchanged, and wait for my approval before running reviews.

    • Read your context and instructions
    • Compiled the paper writing agent

    Illustrative planning-stage response for the synthetic toy-cache draft:

    I found main.tex and main.pdf. I propose three separate initial reviews of the same frozen text: methods and soundness, evidence and evaluation, and significance and positioning. A fourth job will combine their findings after the initial reviews are complete.

    The jobs use your selected, signed-in agent CLI with its default model. Each job receives the manuscript text and the common review form; the initial reviewers do not receive each other’s reviews. Automated jobs cannot inspect figure pixels or page layout. The proposed limit is 180 seconds per attempt, with at most two attempts per job.

    Because this draft has no measured results, the review should distinguish unclear writing from missing evidence. Any proposed experiments will state the claim to test, comparisons, metrics, and completion condition. They will remain research proposals for your discussion and approval.

    Status: awaiting your approval of this review plan. No reviewer jobs run under this plan yet. After the reviews, we will discuss the proposed tasks. Writing changes will be presented as an exact diff for separate approval before application, and builds need their stated approval scope.

    I found main.tex and main.pdf

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

    About this skill

    Work on a research paper through conversation, from early notes and related work to revisions and preparation of the final PDF. Paper Writing Agent helps clarify the argument, explain the method, connect claims to evidence, and improve prose, figures, and tables.

    Install it once for your agent, then open a paper folder and describe what you need. The agent handles project setup while preserving existing manuscript files.

    The review loop brings three perspectives: methods, evidence, and significance. It combines their findings into an assessment and a proposed improvement plan, separating writing changes from questions that need existing evidence, new experiments, or proof.

    You stay in control: approve the review plan before jobs run, select the tasks to pursue, inspect the exact edits before application, and authorize builds within their stated scope. Saved progress supports pausing, resuming, and cancellation.

    Automated CLI reviews assess text; visual inspection is a separate task. Agensi provides the instructions through MCP. The complete local workflow uses the scripts and resources in the MIT-licensed public package: https://github.com/Lemmy00/paper-writing-agent. MCP access alone does not install that runtime.

    Automated CLI reviews assess supplied text; they do not inspect figure pixels or PDF layout. Scientific decisions and acceptance readiness still need author judgment. Antigravity CLI compatibility fixes pass offline tests, but a successful live review remains unconfirmed. Install the complete local runtime separately from the public GitHub package.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 6 days ago

    • Free with an Agensi account

    Requires

    A supported agent: Codex, Claude Code, or Antigravity
    Python 3.10+ for local helpers
    Python
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Browser
    Network Access

    File Scopes

    **/*

    Access stays within the author-approved paper project and selected references. Browser/network access supports literature lookup and approved provider calls. Reviewer CLI jobs receive frozen manuscript text with tools restricted; they do not execute experiments. Proposed manuscript changes require separate diff approval.

    Listed6 days ago

    What's inside

    Frequently Asked Questions

    No. Ask your agent to install the research-paper skill from the public GitHub package once. Then open a paper project and invoke it through chat. The agent handles setup and configuration. Agensi MCP access alone does not install the local runtime.

    It first presents a concrete plan. Reviews run only within approved scope. After review, you discuss and select tasks. Writing tasks produce a proposed diff for separate approval before application. Research tasks and builds require their own applicable scope; review scores do not guarantee acceptance.

    Yes. You choose reference papers, researchers, and style preferences during setup. There are no default professors or researchers. Your choices belong in your paper project, outside the shared skill.

    New in Research & Analysis