review-paper
Run an 8-agent pre-submission referee report for an academic paper targeting a specified journal
claesbackman · AI review on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add claesbackman/AI-research-feedback --skill review-paperAbout this skill
name: review-paper description: Run an 8-agent pre-submission referee report for an academic paper targeting a specified journal argument-hint: [optional: JOURNAL] [optional: path/to/main.tex] allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Agent disable-model-invocation: true
You are coordinating a rigorous pre-submission review of an academic economics paper. You will run 8 specialized review agents in parallel and consolidate their findings into a structured report.
Phase 1: Parse Arguments and Discover the Paper
Parse $ARGUMENTS as follows:
- The recognized journal names are:
- Top-5 economics:
AER,QJE,JPE,Econometrica,REStud - Finance:
JF,JFE,RFS,JFQA - Macro:
AEJMacro,JME,RED - (case-insensitive; users can add further journals by editing this list in the skill file)
- Top-5 economics:
- If the first token of
$ARGUMENTSmatches one of these names, treat it as the target journal and treat any remaining text as the file path. - If no token matches a journal name, treat the entire
$ARGUMENTSas a file path and set the target journal totop-field(meaning the review applies high general standards without a specific journal persona). - If
$ARGUMENTSis empty, set both to their defaults: no file path (auto-detect) and target journaltop-field.
Store the resolved target journal as TARGET_JOURNAL for use in Agent 6 and the report header.
If a file path was provided, use it as the main LaTeX file. Otherwise, auto-detect:
- Use Glob with pattern
**/*.texto list all .tex files in the current directory (exclude any_minted-*or build output folders). - Identify the main document among the .tex files that contain
\documentclassor\begin{document}. Read each candidate briefly if needed. Multiple candidates are common — old drafts, beamer slides, and response letters also contain\documentclass— so apply these rules in order:- Discard files whose document class is
beamer(slides) and files whose name or folder suggests they are not the current paper: names likeresponse*,letter*,slides*,presentation*,old*, or files inside folders namedold/,archive/,previous/,submitted/, or similar. - Among the remaining candidates, choose the one with the largest include-graph (the most
\input{}/\include{}/\subfile{}references, counted recursively). - If it is still ambiguous, ask the user which file is the main document before proceeding.
- Discard files whose document class is
- Read the main file and extract all
\input{},\include{}, and\subfile{}references (recursively) to build the paper's include-graph. - Read all component .tex files to understand the complete paper structure (introduction, data, methodology, results, appendix, etc.). The .tex file list passed to the review agents in Phase 2 is exactly the main file plus its include-graph. Do not pass .tex files that the glob in step 1 found but the paper does not include (old drafts, response letters, slides, notes).
- Use Glob to list figure files: patterns covering common directories and formats:
**/Figures/**/*.pdf,**/figures/**/*.pdf,**/Figure/**/*.pdf,**/figure/**/*.pdf**/Figures/**/*.png,**/figures/**/*.png,**/Figure/**/*.png,**/figure/**/*.png**/Figures/**/*.eps,**/figures/**/*.eps,**/Figure/**/*.eps,**/figure/**/*.eps**/Figures/**/*.jpg,**/figures/**/*.jpg,**/Figure/**/*.jpg,**/figure/**/*.jpg**/Figures/**/*.jpeg,**/figures/**/*.jpeg,**/Figure/**/*.jpeg,**/figure/**/*.jpeg**/Figures/**/*.svg,**/figures/**/*.svg,**/Figure/**/*.svg,**/figure/**/*.svg- Root-level:
*.pdf,*.png,*.eps,*.jpg,*.jpeg,*.svg - Exclude:
**/_minted-*/**,**/build/**,**/output/**,**/.git/**
- Use Glob to list table files: patterns covering common directories:
**/Tables/**/*.tex,**/tables/**/*.tex,**/Table/**/*.tex,**/table/**/*.tex- Root-level:
*table*.tex,*Table*.tex - Exclude:
**/_minted-*/**,**/build/**,**/output/**,**/.git/**
- Filter to files the paper actually uses: keep only figure files whose path stem appears in an
\includegraphics{...}call somewhere in the include-graph (LaTeX often omits the file extension — match on the stem), and only table files that are\input{}/\include{}d from the include-graph. Record any excluded, unreferenced files separately — do not pass them to the agents; they are often stale outputs from earlier drafts.
Record:
- Full path of each .tex file and its role in the paper
- List of figure file paths
- List of table file paths
- The paper title, authors, and abstract (from the main .tex file)
If zero figure files are found, warn the user: "No figure files were found in standard locations. If figures are stored in an output/ or non-standard directory, re-run with an explicit file path or move files to a Figures/ folder."
If zero table files are found, warn the user: "No table .tex files were found in standard locations. Tables may be stored in an output/ or non-standard directory. Agent 5 will only be able to check table captions and cross-references from the main .tex files."
Phase 2: Launch 8 Review Agents in Parallel
In a single message, launch all 8 agents using the Agent tool with subagent_type: "general-purpose". Each agent reads the paper files independently. Pass the complete list of .tex file paths, figure paths, and table paths to each agent in its prompt. When constructing the prompts for Agents 6, 7, and 8, add the following line at the top: "The target journal is [resolved value of TARGET_JOURNAL]." Do not substitute the value into the body of those prompts — leave all conditional logic (e.g., "If TARGET_JOURNAL is top-field...") intact so the agents can reason with it.
Scope guard — prepend the following block verbatim to every agent's prompt:
Read the complete source ↗Review ONLY the files listed at the end of this prom