Respond to Referee Skill
SkillDocs & knowledgeDraft response letter text and LaTeX edits for referee points, or generate a complete reply document
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Respond to Referee Skill skill
What this skill tells your AI
The instructions your AI receives, as published by alexander-m-dickerson/ai-asset-pricing in .claude/skills/respond-to-referee/SKILL.md and read by ahel’s review.
Structured workflow for addressing referee comments. Operates in two modes:
- Single-point mode: Draft a response paragraph + LaTeX edits for one referee point
- Full-reply mode: Generate a complete standalone reply LaTeX document
Examples
/respond-to-referee referee2.md point 3-- address a specific point from a referee file/respond-to-referee referee2.md-- generate complete reply to all points in the file/respond-to-referee identification-- address a point by topic keyword
Input
The user provides:
- A referee report file (markdown or text)
- Either a specific point number, a topic keyword, or no argument (full reply)
Core Principles
These principles, distilled from journal editor guidelines and published advice (Noble 2017, PLOS Comp Bio; Review of Finance "Tips for Authors"), govern all referee responses.
Tone
- Grateful but not obsequious: "We thank the referee for this suggestion" (good) vs. "We are deeply grateful for this invaluable insight" (too much)
- Accept blame for misunderstandings: If the referee misread something, it is our exposition failure. "We realize our original text was ambiguous and have revised it as follows..."
- Never dismissive: No bare "we respectfully disagree." Every disagreement must be backed by evidence, data, or a reference.
- Direct answers first: Start each response with what you did ("We have added...", "We agree and now show..."), then explain why.
- Remember the audience: You are writing to the editor, not (only) the referee.
Structure
- Respond to every point: No exceptions, including minor ones.
- Self-contained responses: Quote or paraphrase the revised manuscript text directly in the letter.
- Reference by section name: Not page numbers (which shift between drafts).
- Group related points: When two comments address the same underlying issue, respond jointly.
Substance
- Do what the referee asks, even if you disagree: Run the requested analysis, report results in the letter, then explain why you believe the main text should differ.
- Don't over-revise: Restrict changes to what is requested.
- Address general criticisms globally: If the referee cites two examples, fix the problem paper-wide.
Common Mistakes to Avoid
| Mistake | Why it fails |
|---|---|
| Bare "we respectfully disagree" | Editor has no basis to overrule |
| Sarcasm or condescension | Poisons the review relationship |
| Fixing only the specific examples cited | Signals you missed the general point |
| Long defensive paragraphs before stating what you did | Buries the lede |
| Page-number references | Numbers shift between drafts |
| Over-revising (large unrequested changes) | Creates new attack surface |
Em-dashes (---) in prose | AI hallmark; see academic-writing.md |
| Hard-banned words (delve, crucial, etc.) | LLM tells; see academic-writing.md Section 1 |
| Hedge words (somewhat, quite, very, arguably) | Weaken claims; give magnitudes instead |
Mode 1: Single-Point Workflow
Step 1: Load Context
- Read the referee report file
- Read the project's
CLAUDE.mdfor paper state and claims - Read
.claude/rules/academic-writing.mdfor style rules - Identify the specific referee point
Step 2: Map to Affected Sections
- Identify which section(s) in the
.texfile are affected - Read the affected passage(s)
- Check if the point has already been partially addressed
Step 3: Draft Response Letter Paragraph
\item[\textbf{Referee:}] \textit{``[abbreviated quote of referee comment]''}
\item[Reply:] [We thank the referee for... / We agree that...]
[1-2 sentences explaining what we did and where]
[If helpful: quote the revised text in a \begin{quote} block]
[1-2 sentences explaining the rationale]
Step 4: Draft LaTeX Edits
- Write proposed edits using the paper's terminology
- Follow all rules from
academic-writing.md - Show old text -> new text for each edit
Step 5: Consistency Check
Verify proposed edits do not:
- Contradict claims in other sections
- Change quantitative results that appear elsewhere
- Introduce terminology violations
- Break cross-references
Step 6: Output
REFEREE RESPONSE: Point [N] -- [topic]
======================================
REFEREE SAID:
[brief summary]
RESPONSE LETTER TEXT:
[draft LaTeX paragraph]
PROPOSED EDITS:
File: [filename]
Location: Section [key], ~line [N]
OLD: [existing text]
NEW: [proposed replacement]
CONSISTENCY NOTES:
- [any cross-section impacts]
STATUS: [Ready / Needs human review on X]
Mode 2: Full-Reply Workflow
Step 1: Load Context
- Read the referee report file (all points)
- Read the project's
CLAUDE.mdandacademic-writing.md
Step 2: Plan the Reply
- Enumerate all referee points (substantive + minor)
- Check which have already been addressed
- Group related points
- Flag any points where paper edits are still needed vs. reply-only
Step 3: Create Reply Document
- Create
_replies/reply_referee{N}.texusing the LaTeX template below - Point-by-point responses with
\item[\textbf{Referee:}]/\item[Reply:] - Opening and closing paragraphs
Step 4: Compile and Verify
- Run pdflatex + bibtex + pdflatex + pdflatex (paths from canonical local state reported by
tools/bootstrap.py audit, or a repo-root compatibility shim if present) - Check for undefined citations or references
- Read through for tone, completeness, accuracy
Step 5: Output
Report the compiled PDF location, page count, and any issues.
LaTeX Template (for full replies)
\documentclass[12pt,letterpaper]{article}
\usepackage[top=1.0in, bottom=1.0in, left=0.90in, right=0.90in]{geometry}
\usepackage{amsfonts,amsmath,amssymb}
\usepackage{setspace,titlesec,xcolor,booktabs,enumerate}
\usepackage{natbib}
\usepackage{hyperref}
\renewcommand{\baselinestretch}{1.10}
\titleformat{\section}{\centering\large\bfseries}{\thesection.}{1em}{}
\renewcommand{\thesection}{\Roman{section}}
\begin{document}
\noindent \textbf{{\large Reply to Referee [N] for ``{Paper Title}''}}
\medskip
\noindent Manuscript [ID]
\bigskip
\noindent [Opening paragraph: thank referee, brief overview of revision scope]
\section{Response to the Referee's comments}
\noindent
\begin{enumerate}
\item[\textbf{Referee:}] \textit{``[quoted comment]''}
\item[Reply:] [response]
% ... repeat for each point ...
\end{enumerate}
\bigskip
\noindent [Closing paragraph]
{\footnotesize
\singlespacing
\setlength{\bibsep}{6pt}
\bibliographystyle{jf}
\bibliography{../latex/references}
}
\end{document}
Signals
- GitHub stars
- 59
- Forks
- 10
- Last commit
- Apr 2026
Advanced
- Catalog kind
- skill
- Gateway key
respond-to-referee- Source
- github.com/alexander-m-dickerson/ai-asset-pricing