agentleFS
Xuzhen-LiXuzhen-Li/literature-review-skills/skills/reviewer-response/SKILL.md

reviewer-response

Skill2 starsChanged 35 days agoMITView on GitHub

Numbers every editor and reviewer comment, plans the action, and drafts a point-by-point response that cites the page and lines of each real change. Use when the user says "response to reviewers", "decision letter", "point by point", "R1.1", or "revise the manuscript". Do not use to invent comments, to claim a sentence was changed when it was not, or to submit the revision.

  • Installs packages

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 7 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Read the letter. Split it into separate requests.

    line 35
  2. Very specific:

    Write <project>/14-revision/action_plan.md before you edit the manuscript.

    line 37
  3. Very specific:

    Write <project>/14-revision/response_table.csv with this header:

    line 43
  4. Very specific:

    Write <project>/14-revision/response.md from references/response_template.md.

    line 51
  5. Distinctive:

    Every comment in the letter has one ID, and the IDs have no gaps.

    line 92
  6. Very specific:

    Every "we have revised" (and the other revision phrases) sits next to p.

    line 93
  7. Distinctive:

    Do not invent a reviewer comment, a compliment, or a change.

    line 99
  8. Distinctive:

    Do not write "we have addressed all comments" as a substitute for the point-by-point table.

    line 100
  9. Distinctive:

    Do not add data, a panel, or a citation that is not in the evidence table.

    line 101
  10. Distinctive:

    Do not contact the reviewers outside the journal's system.

    line 102
  11. Distinctive:

    Stop. The user uploads the revised manuscript, the response docx, and any changed figures.

    line 107

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Reviewer response

Opening

On first use in a project, say this verbatim:

I am reviewer response. I answer each comment from the decision letter and I cite only changes that were made.

When to use

The user has a decision letter and wants a revision, or the letter invites a response. The letter is saved verbatim before you number anything. You do not write a response to a comment that is not in that file.

Inputs (paths) / Preconditions

  • <project>/14-revision/decision_letter.md, .txt, or .pdf — pasted or saved by the user, unchanged. Read whichever of those three exists. Do not rename a .txt or .pdf the user already saved.
  • The manuscript file you will actually change, usually <project>/10-polish/manuscript.md, plus the docx the journal will receive.
  • <project>/evidence/evidence_table.csv if a new citation is required. A new citation still has to be a real row. If it is not in the table, stop and go back to paper-deep-reading. Do not invent one to satisfy a reviewer.
  • references/tone_guide.md and references/response_template.md

If the letter is a rejection and does not invite revision, write <project>/14-revision/decision_note.md saying so, quote no comments you do not have, and stop. An appeal is a separate request. Read the journal's appeal policy first. Do not draft an appeal in this skill unless the user asks after that page has been opened.

Procedure (numbered steps)

  1. Read the letter. Split it into separate requests. Do not merge two requests into one number, and do not drop a criticism because it is inconvenient.
  2. Number them in the letter's order. Editor comments are E.1, E.2, with no gaps. Reviewer 1 is R1.1, R1.2. Reviewer 2 is R2.1. If there is a reviewer 2, there is a reviewer 1. Editor comments come first when the letter has them.
  3. Write <project>/14-revision/action_plan.md before you edit the manuscript. One heading per ID and one action bullet: revised, declined, or clarified. Show the plan to the user.
  4. After the user accepts the plan, edit the manuscript. A revised action must change the file. A declined action must not. A clarified action changes the file only when you also use a revision phrase; otherwise it points the reviewer to text that is already there.
  5. Turn on line numbers in the file you will submit before you write locations. Record both locators:
    • Source locator, for the checker: 10-polish/manuscript.md lines 40-44
    • Letter locator, which the reviewer reads: p. 3, lines 40-44 The page and line numbers are read off that file. Do not estimate them. If the docx has no line numbers yet, do not use the phrase "we have revised".
  6. Write <project>/14-revision/response_table.csv with this header:
comment_id,action,location,change_quote,response

location is none when you did not edit. Otherwise it is the source locator. change_quote is the new text, at least a clause, and it must occur inside those lines. response is the paragraph the reviewer will read. For a revision it contains both a revision phrase (we have revised, we have added, we have changed, we have deleted, or we have clarified) and the page-line pattern p. 3, lines 40-44. It does not contain a project path such as 10-polish/.

  1. Write <project>/14-revision/response.md from references/response_template.md. Each section heading is the comment ID. The body contains the response cell verbatim. Follow references/tone_guide.md. Do not write "the reviewer misunderstood" unless you can point to the sentence they misread, and even then answer the request.
  2. Run:
python3 <suite>/skills/reviewer-response/scripts/check_response.py \
  --table <project>/14-revision/response_table.csv \
  --letter <project>/14-revision/response.md \
  --plan <project>/14-revision/action_plan.md \
  --root <project>

The command fails if a revision phrase has no page and line, if the quoted new text is not in the cited lines, if a declined comment says you revised, or if the plan and the letter disagree. Fix the manuscript or the letter. Do not weaken the checker to make it pass.

  1. Export the letter:
pandoc "<project>/14-revision/response.md" \
  --from=markdown \
  --to=docx \
  --output="<project>/14-revision/response_to_reviewers.docx"

Pandoc's manual is https://pandoc.org/MANUAL.html. If pandoc is missing, use the same fallback rule as manuscript-polishing (skills/manuscript-polishing/references/export_docx.md). Do not pip install.

  1. Append YYYY-MM-DD | reviewer-response | response checked against the revised file | agent only when the command prints OK response. The user resubmits.

Human gate(s)

Revision plan: do not edit the manuscript until the user accepts action_plan.md. If they refuse an action, change the plan, do not silently revise. Resubmission is the user's act. You do not log in.

Outputs (exact paths + formats)

  • <project>/14-revision/action_plan.md
  • <project>/14-revision/response_table.csv
  • <project>/14-revision/response.md
  • <project>/14-revision/response_to_reviewers.docx
  • The revised manuscript, with the quoted changes present at the cited lines

Quality checklist

  • The checker prints OK response.
  • Every comment in the letter has one ID, and the IDs have no gaps.
  • Every "we have revised" (and the other revision phrases) sits next to p. N, lines N-N, and the new wording is in the file at the source locator.
  • Declined comments say what you did not change and why, with an evidence ID when the reason is scientific.
  • The docx the journal receives does not contain 10-polish/ or other project paths.

Guardrails

  • Do not invent a reviewer comment, a compliment, or a change.
  • Do not write "we have addressed all comments" as a substitute for the point-by-point table.
  • Do not add data, a panel, or a citation that is not in the evidence table.
  • Do not contact the reviewers outside the journal's system.
  • A second round uses a new letter file and a new table. Do not reuse locations from the previous round after the file has moved.

Next step

Stop. The user uploads the revised manuscript, the response docx, and any changed figures. If another letter arrives, run this skill again on that letter.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.