Skip to content
Get packs

Auto Paper Improvement Loop

Repeatedly review a generated paper with an external model, fix it, and recompile, for two rounds.

Source Download skill
Repository
wanshuiyin/Auto-claude-code-research-in-sleep
Stars
18,482
Licence
MIT
Files
1 file

What it does

After a paper has been generated and compiled, this skill makes the agent send it to an external AI reviewer (GPT-6-Astra via Codex MCP) for critique, implement the suggested writing fixes, and recompile the paper. It runs two such review-fix-recompile rounds automatically, logging each round to a file in the paper folder. It targets writing quality such as theoretical inconsistencies, overclaims, missing content and presentation, not running new experiments.

Great for

  • Researchers who have a compiled draft from the paper-writing workflow and want a first polishing pass
  • Users who want an automated, logged review cycle rather than manual back-and-forth with a reviewer

Skip it if

  • Papers that need new experiments or data, because this skill only edits writing; use the research loop instead
  • Final submissions without human reading, since the reviewer and the edits are automated and may introduce errors or overclaims that need human checking
  • Setups without the external reviewer tools configured, since the review step will not run

Try asking

  • “Improve the paper in paper/ with two rounds of automated review and fixes”
  • “auto improve my compiled draft in ./paper with a style reference from arXiv:2401.00001”
  • “Run the paper improvement loop on paper/ with an edit whitelist limiting changes to the sections folder”
Step-by-step guide: how to use Auto Paper Improvement LoopInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    auto-paper-improvement-loop.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip auto-paper-improvement-loop.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip auto-paper-improvement-loop.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Start with a paper that has already been compiled and reviewed by you, so the loop refines rather than rescues a broken draft.
  • Use the edit whitelist option to restrict which files the agent may change, and review the PAPER_IMPROVEMENT_LOG.md afterwards.
  • Do not schedule this skill with /loop or cron; it already loops internally and fresh reviews are the point.

What's inside

  • SKILL.md

Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.