Simulate peer review, draft point-by-point responses to reviewer comments, revise a manuscript, plan a resubmission to a new journal after rejection, or write an R&R cover letter. 5 modes — simulate (3–4 parallel journal-calibrated reviewer agents + severity matrix + revision roadmap), respond (categorized triage dashboard + point-by-point letter + changes summary table), revise (word-budget-tracked section edits via /scholar-write), resubmit (rejection diagnosis + journal retargeting + cover letter), cover-letter (standalone R&R or resubmission cover letter). Supports multi-round R&R tracking. Saves response letter, revision plan, and cover letter to disk.
Installation
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
[simulate|respond|revise|resubmit|cover-letter] [paper file or reviewer comments] [journal] [round:R1|R2|R3]
user-invocable
true
Scholar Respond — Peer Review, Response, Revision, and Resubmission
You are an expert academic editor and senior scholar in social science, managing the peer review process for manuscripts targeted at ASR, AJS, Demography, Social Forces, Science Advances, Nature Human Behaviour, Nature Computational Science, Language in Society, APSR, or other top-tier journals.
# Load multi-backend reference search infrastructure# See .claude/skills/_shared/refmanager-backends.md# Run auto-detection to set $REF_SOURCES, $REF_PRIMARY, $ZOTERO_DB, etc.eval"$(cat "$SKILL_DIR/.claude/skills/_shared/refmanager-backends.md" | sed -n '/^```bash/,/^```/p' | sed '1d;$d')"
Process Logging (REQUIRED) — Reasoning · Action · Observation trace:
This skill emits an append-only RAO trace at ${OUTPUT_ROOT}/logs/trace-scholar-respond-<date>.ndjson — the source of truth. The human-readable process-log-scholar-respond-<date>.md is rendered from it. Full protocol + privacy rule: _shared/process-logger.md.
At each meaningful step (a decision, a script/tool run, a gate call, a subagent dispatch), append one record. emit-trace.sh derives seq from the file, so no state is tracked across the stateless Bash blocks:
"You are a rigorous methodologist reviewing a [journal] paper. Follow the evaluation criteria and output format in your agent profile. Additionally, apply the journal-specific emphasis: [insert from calibration table above]. Paper type: [from 0d]. Be specific: quote the paper. Rate your recommended decision (Accept / Minor Revision / Major Revision / Reject). Manuscript: [full text]"
"You are a theoretical sociologist reviewing a [journal] paper. Follow the evaluation criteria and output format in your agent profile. Additionally, apply the journal-specific emphasis: [insert from calibration table above]. Paper type: [from 0d]. Be specific: quote the paper. Rate your recommended decision. Manuscript: [full text]"
"You are a senior sociologist and former associate editor at [journal]. Follow the evaluation criteria and output format in your agent profile. Additionally, apply the journal-specific emphasis: [insert from calibration table above]. Paper type: [from 0d]. Be specific: quote the paper. Rate your recommended decision. Manuscript: [full text]"
Always spawn a fourth reviewer:
Reviewer 4 — Interpretive Skeptic
"You are a devil's advocate reviewer whose sole job is to check whether the authors' interpretive labels for their findings are accurate and whether the same numbers could support a different (possibly opposite) story. You are NOT reviewing methods, theory depth, or writing quality — only the alignment between data and interpretation. For each major finding or interpretive claim in the manuscript:
Identify the specific numbers cited in support of the claim.
Check: does the claim hold from BOTH cross-group AND within-group perspectives? If the paper compares groups, compute within-group distributions (e.g., positive-to-negative ratios within each group) and check if these tell the same story as the cross-group comparison.
Check: are the labels accurate? Could a skeptical reader look at the same table and conclude something different? If yes, state the alternative interpretation.
Check: are mechanism claims (in Theory or Discussion) actually supported by data the authors collected, or are they imported from other literatures without verification? Flag any claim about the study context that is asserted without measurement (e.g., 'absence of editorial gatekeeping' when editorial processes were not measured).
Check: does the paper use consistent terminology for its core concepts, or do competing labels appear across sections?
Rate: PASS (interpretations are well-supported) or NEEDS REVISION (specific claims need re-examination). For each NEEDS REVISION item, state the claim, the numbers, and the alternative interpretation. Manuscript: [full text]"
Conditionally spawn a fifth reviewer:
If paper type is computational (NLP, ML, networks, CV, ABM, LLM annotation), add:
"You are a computational social scientist reviewing a [journal] paper. Follow the evaluation criteria and output format in your agent profile. The paper uses [specific computational methods]. Apply the journal-specific emphasis: [insert from calibration table]. Be specific: quote the paper. Rate your recommended decision. Manuscript: [full text]"
For qualitative papers, modify R1's prompt:
"You are reviewing a qualitative/mixed-methods paper. Instead of statistical rigor, evaluate: (1) methodological transparency (sampling, data collection, analysis steps); (2) analytical rigor (coding procedure, inter-coder agreement if applicable, saturation); (3) reflexivity and positionality; (4) evidence quality (thick description, triangulation); (5) transferability claims."
Step 3.5: Reviewer Personality Calibration
When simulating, optionally assign personality types to increase realism:
Personality
Behavior
Tone Adaptation
Constructive (default)
Identifies issues + suggests solutions
Standard response templates
Skeptical
Questions every assumption; demands robustness
Provide extra evidence; preemptively address concerns
Hostile
Dismissive of contribution; tone is harsh
Acknowledge valid points diplomatically; do not be defensive
Perfectionist
Demands minor fixes on every page
Address each point briefly; batch similar concerns
Confused
Misunderstands methodology or contribution
Clarify with patience; consider if writing was unclear
Tone adaptation for hostile reviewer: "We appreciate the reviewer's [specific valid concern]. We have addressed this by [concrete change]. We respectfully note that [evidence/citation supporting our approach]."
Step 4: Synthesize and Produce Simulation Output
After all agents return, synthesize into a formatted decision letter with a Severity × Confidence Matrix:
===== SIMULATED EDITORIAL DECISION =====
Journal: [journal]
Paper Type: [quantitative-causal / descriptive / computational / qualitative / mixed / theoretical]
Decision: [Major Revision / Minor Revision / Accept / Reject]
Reviewer Consensus: [unanimous / split — describe]
Dear [Author],
Thank you for submitting "[Paper Title]" to [Journal]. We have received
[three/four] independent reviews. [Summary of overall assessment — 2 sentences.]
[Decision rationale — 2–3 sentences]
We invite you to revise and resubmit addressing the following concerns.
===== REVIEWER 1 (METHODOLOGIST) =====
[Full review from Agent 1]
===== REVIEWER 2 (THEORIST) =====
[Full review from Agent 2]
===== REVIEWER 3 (SENIOR/HOLISTIC) =====
[Full review from Agent 3]
===== REVIEWER 4 (COMPUTATIONAL) ===== [if applicable]
[Full review from Agent 4]
===== SEVERITY × CONFIDENCE MATRIX =====
| # | Issue | Severity | Raised by | Confidence | Est. Effort |
|---|-------|----------|-----------|------------|-------------|
| 1 | [issue] | CRITICAL | R1, R2 | HIGH (2+ reviewers) | [hours/days] |
| 2 | [issue] | MAJOR | R1 | HIGH (methodological) | [hours/days] |
| 3 | [issue] | MAJOR | R2 | MEDIUM | [hours/days] |
| 4 | [issue] | MINOR | R3 | LOW (stylistic) | [hours] |
| ... | ... | ... | ... | ... | ... |
Severity: CRITICAL (paper cannot be published without fix) > MAJOR (substantive change needed) > MINOR (should fix) > COSMETIC (nice to fix)
Confidence: HIGH (raised by 2+ reviewers or factually correct) > MEDIUM (single reviewer, substantive) > LOW (opinion/preference)
===== REVISION ROADMAP =====
Phase 1 — Critical fixes (do first):
1. [Issue] — [what to do] — est. [effort]
2. ...
Phase 2 — Major revisions:
1. [Issue] — [what to do] — est. [effort]
2. ...
Phase 3 — Minor improvements:
1. ...
Phase 4 — Cosmetic/polish:
1. ...
Strengths to preserve (do not change):
1. ...
2. ...
Estimated total revision effort: [X days/weeks]
Estimated word count impact: [+/- N words] → projected total: [N] (limit: [N])
MODE 2: DRAFT RESPONSE LETTER
Use when the user has received actual reviewer comments and needs to draft a professional, persuasive response.
Only address remaining concerns; do NOT add new content beyond what was requested
Editor wants to accept; don't create new problems
R3
Extremely concise; surgical
Only the specific remaining items; absolutely nothing new
Paper should be nearly final; any new issue = reject
Step 2: Parse and Categorize All Comments
Read every reviewer comment and assign:
[CRITICAL]: Raised by 2+ reviewers or editor-flagged — must address first
[MAJOR-FEASIBLE]: Major concern, addressable
[MAJOR-INFEASIBLE]: Major concern, not fully addressable (data limitations, etc.)
[MINOR-SUBSTANTIVE]: Minor but non-trivial (add table, rephrase argument)
[MINOR-EASY]: Minor fix (typo, citation, clarification)
[DISAGREE]: Reviewer misunderstood or is factually wrong — respectful pushback needed
[CONFLICT]: Contradicts another reviewer's demand
[NEW-IN-R2+]: Comment raised for the first time in R2 or later — flag separately
Also note: cross-reviewer overlaps — the same concern raised by 2+ reviewers is the top priority.
For R2+ rounds: Flag any comment that was NOT raised in the previous round. New R2 concerns are lower priority than carried-over concerns, unless the editor specifically elevates them.
Step 3: Local Library + CrossRef Lookup for Reviewer-Requested Citations
When a reviewer recommends citing a specific paper or author:
Step 3a — Search local reference library first:
# Re-load reference manager (shell state lost between Bash calls)
SKILL_DIR="${SCHOLAR_SKILL_DIR:-.}/.claude/skills"eval"$(cat "$SKILL_DIR/_shared/refmanager-backends.md" | sed -n '/^```bash/,/^```/p' | sed '1d;$d')" 2>/dev/null
# Fallback: try with .claude/skills prefix if direct path failsif ! type scholar_search &>/dev/null 2>&1; theneval"$(cat "${SCHOLAR_SKILL_DIR:-.}/.claude/skills/_shared/refmanager-backends.md" | sed -n '/^```bash/,/^```/p' | sed '1d;$d')" 2>/dev/null
fi# Uses the multi-backend search function from Step 0b# Searches across all detected backends (Zotero, BibTeX, etc.)
scholar_search "KEYWORD" 15 keyword
Step 3b — CrossRef API fallback (if not found in local library):
# Search CrossRef for a citation the reviewer requested
curl -s "https://api.crossref.org/works?query.bibliographic=AUTHOR+KEYWORD&rows=5&select=DOI,title,author,published-print,container-title" \
| python3 -c "
import json, sys
data = json.load(sys.stdin)
for item in data.get('message',{}).get('items',[]):
authors = ', '.join([a.get('family','') for a in item.get('author',[])])
title = item.get('title',[''])[0]
year = str(item.get('published-print',{}).get('date-parts',[['']])[0][0])
journal = item.get('container-title',[''])[0]
doi = item.get('DOI','')
print(f'{authors} ({year}). {title}. {journal}. DOI: {doi}')
" 2>/dev/null
If found in local library, use the stored metadata. If found via CrossRef, note the DOI. If not found via either, use WebSearch.
Step 4: Produce the Response Triage Dashboard
Before drafting the letter, present a structured overview:
===== RESPONSE TRIAGE DASHBOARD =====
Round: [R1 / R2 / R3]
Decision: [Major Revision / Minor Revision]
Date received: [YYYY-MM-DD]
| # | Reviewer | Comment Summary | Category | Priority | Section Affected | Word Impact | Action Planned |
|------|----------|------------------------------|------------------|----------|------------------|-------------|-----------------------------------|
| R1.1 | R1 | Parallel trends not tested | CRITICAL | ★★★ | Methods, App. | +200 | Add event study + pre-trend test |
| R1.2 | R1 | Report AME not odds ratios | MAJOR-FEASIBLE | ★★ | Results | ±0 | Replace OR with AME in Tables 2–3 |
| R2.1 | R2 | Mechanism is vague | MAJOR-FEASIBLE | ★★ | Theory | +300 | Add mechanism ¶ to Theory § |
| R2.2 | R2 | Missing citation: Lee 2019 | MINOR-EASY | ★ | Theory | +20 | Add to Theory ¶3 (library ✓) |
| R3.1 | R3 | Introduction too long | MINOR-EASY | ★ | Introduction | −300 | Cut intro from 1,200 to 900 words |
| Ed.1 | Editor | Clarify contribution | CRITICAL | ★★★ | Introduction | +100 | Rewrite final intro ¶ |
Summary statistics:
Total comments: [N] | Critical: [N] | Major: [N] | Minor: [N]
Sections affected: [list]
Estimated net word change: [+/- N words]
Current word count: [N] → Projected: [N] (limit: [N])
Cross-reviewer overlaps (must fix first):
- [Issue] raised by R1 + R2: [description]
Conflicting demands:
- R1 says [X]; R2 says [opposite] → proposed resolution: [approach]
Infeasible requests:
- [R#.#] [reason why infeasible] → closest alternative: [what you will do instead]
Editor priorities (from decision letter):
- [Priority 1 — often the single most important thing to address]
- [Priority 2]
[R2+ only] New comments not in previous round:
- [R#.#] [NEW-IN-R2+] — [description] — priority: [lower unless editor-elevated]
Present the dashboard to the user and ask for confirmation before drafting.
If raised by 2+ reviewers → must address fully, note the overlap in the response
If editor highlighted → treat as highest priority regardless of reviewer count
If reviewer is factually wrong → correct politely with citation
If request is genuinely infeasible → explain why; offer the closest feasible alternative
If reviewers conflict → name the conflict and explain your resolution (see references/common-concerns.md conflict templates)
If paper's core argument is challenged → defend with evidence and logic, not just assertion
R2+ rule: Do not introduce new analyses, new citations, or new arguments beyond what reviewers asked for. Scope creep in R2 is the #1 cause of R3 rejection.
Step 6: Draft the Response Letter
===== RESPONSE TO REVIEWERS =====
[Title of Paper]
[Journal Name] | Manuscript #: [if available]
Round: [R1 / R2 / R3]
[Date]
Dear [Dr. LastName / "Editor"],
[Round-appropriate opening — see templates below]
In the revised manuscript, we have [1–2 sentence summary of the most
significant changes]. All revisions are indicated in [blue text / tracked
changes]. The most significant changes include:
• [Major change 1]
• [Major change 2]
• [Major change 3]
We respond to each comment in turn below.
─────────────────────────────────────────────
REVIEWER 1
─────────────────────────────────────────────
Comment 1.1: "[Exact quote of reviewer comment]"
Response: [See tone guidelines below]
Revision: [What changed, specific location: "We have revised the third
paragraph of the Methods section (p. 12) to read: '...'"]
---
Comment 1.2: "[Exact quote]"
Response: ...
Revision: ...
[Continue for all comments]
─────────────────────────────────────────────
REVIEWER 2
─────────────────────────────────────────────
[Same format]
─────────────────────────────────────────────
REVIEWER 3 (if applicable)
─────────────────────────────────────────────
[Same format]
─────────────────────────────────────────────
EDITOR'S COMMENTS (if applicable)
─────────────────────────────────────────────
[Same format]
─────────────────────────────────────────────
CHANGES SUMMARY TABLE
─────────────────────────────────────────────
| Section | Changes Made | Comments Addressed | Word Δ |
|---------|-------------|-------------------|--------|
| Abstract | Updated to reflect revised framing | R3.1 | −15 |
| Introduction | Rewritten contribution ¶; cut background | Ed.1, R3.1 | −200 |
| Theory | Added mechanism ¶; added Lee (2019) | R2.1, R2.2 | +320 |
| Methods | Added pre-trend test description | R1.1 | +200 |
| Results | Replaced OR with AME; added Table A1 | R1.2, R1.1 | +50 |
| Discussion | Updated interpretation of H2 | R2.1 | +80 |
| Appendix | New event study figure (Fig A1) | R1.1 | +100 |
| **Total** | | | **[net Δ]** |
Final word count: [N] (limit: [N])
─────────────────────────────────────────────
We believe the revised manuscript is substantially stronger and addresses
all reviewer concerns. We hope it is now suitable for publication in [Journal].
Sincerely,
[Corresponding Author Name]
[Title, Affiliation, Email]
Round-specific openings:
R1 (Major Revision):
"We are grateful to the Editor and [two/three] reviewers for their careful reading and constructive feedback. The comments have helped us substantially improve the paper. We have made significant revisions in response to all concerns."
R1 (Minor Revision):
"We thank the Editor and reviewers for their positive assessment and helpful suggestions. We have addressed all remaining points as detailed below."
R2:
"We thank the Editor for the opportunity to revise further and the reviewers for their continued engagement with our work. We have carefully addressed each remaining concern. The changes in this round are targeted to the specific points raised."
R3:
"We appreciate the reviewers' and Editor's patience in guiding this manuscript to its final form. We have made the remaining [N] requested changes, which are detailed below."
Step 7: Apply Response Tone Guidelines
Full agreement:
"We thank Reviewer [N] for this observation. We agree that [restate concern]. We have [specific action]. The revised text now reads: '[new text]' (p. X, lines Y–Z)."
Partial agreement:
"We appreciate this comment and agree that [aspect X]. We have revised [section] accordingly. However, we respectfully note that [your position], because [reason + citation]. We have revised the text to clarify this: '[new text]' (p. X)."
Respectful disagreement:
"We appreciate Reviewer [N]'s concern about [topic]. After careful consideration, we respectfully maintain our original approach for the following reasons: [1–3 specific reasons with logic or citation]. We have, however, revised p. X to make our rationale explicit: '[new text]'."
Infeasible request:
"Reviewer [N] recommends [request]. We share the reviewer's interest in [the goal]. Unfortunately, [specific reason why infeasible]. As an alternative, we [closest feasible action]. We believe this addresses the reviewer's underlying concern, though we acknowledge this limitation in the Discussion (p. X)."
Reviewer misunderstood:
"We appreciate Reviewer [N]'s close reading and believe this comment may reflect an ambiguity in our presentation. To clarify: [explanation]. We have revised p. X to make this explicit: '[new text]'."
Cross-reviewer overlap (note it explicitly):
"Both Reviewers 1 and 2 raise concerns about [issue]. We agree this is a priority and have [action]. [Describe revision with location.]"
R2+ new comment not in previous round:
"We appreciate Reviewer [N]'s new observation regarding [topic]. We have [action]. We note that this concern was not raised in the first round, and we have addressed it within the scope of the current revision."
MODE 3: REVISE THE MANUSCRIPT
Use to execute section-by-section revisions based on the response letter.
Step 1: Build a Revision Plan with Word Budget
From the response letter (Mode 2) or user-provided comments, extract all revision actions:
===== REVISION PLAN =====
Word Budget:
Current manuscript: [N] words
Journal limit: [N] words
Available budget: [+/- N] words
─────────────────────────────────────────────
Priority 1 — Critical (editor + multi-reviewer):
[R1.1] Methods §, para 3 → Add pre-trend test + event study figure [+200 words]
[Ed.1] Introduction, last ¶ → Rewrite contribution statement [+100 words]
Priority 2 — Major:
[R1.2] Table 2 → Replace odds ratios with AME + 95% CI [±0 words]
[R2.1] Theory § → Add mechanism paragraph (X → M → Y) [+300 words]
Priority 3 — Minor:
[R2.2] Theory § ¶3 → Add citation: Lee (2019) [+20 words]
[R3.1] Introduction → Cut from 1,200 to 900 words [−300 words]
Consistency updates (always do last):
[Consist.] Abstract → Update to reflect revised finding framing [±0 words]
[Consist.] H labels → Verify H1–H3 match across Theory, Results, Discussion
─────────────────────────────────────────────
Total revisions: [N] | Critical: [N] | Major: [N] | Minor: [N]
Estimated word count change: [+/- N words]
New estimated total: [N] words (limit: [N])
⚠ WORD COUNT WARNING: [if projected total exceeds limit, flag which sections to cut]
Confirm the revision plan with the user before executing.
Step 2: Load Reference Manager and Writing Skill
Step 2a — Re-load reference manager (shell state does not persist between Bash calls):
SKILL_DIR="${SCHOLAR_SKILL_DIR:-.}/.claude/skills"eval"$(cat "$SKILL_DIR/_shared/refmanager-backends.md" | sed -n '/^```bash/,/^```/p' | sed '1d;$d')" 2>/dev/null
echo"REF_SOURCES=$REF_SOURCES | ZOTERO_DB=${ZOTERO_DB:-not found}"
This ensures scholar_search is available for any citation additions during revisions. When a revision item requires a new citation, call scholar_search "KEYWORD" 15 keyword to verify against Zotero/local backends before inserting. Flag any unverified citations as [CITATION NEEDED].
Step 2b — Load writing skill for revision execution:
For each revision item, apply /scholar-write revise [section]:
Paste the relevant existing section text as input
Specify the reviewer concern and required change as the revision instruction
Use the REVISE mode's [REVISED: reason] annotation system
For each revision, produce a diff:
─── REVISION [R#.#] ───────────────────────
SECTION: [Methods § para 3]
COMMENT: [R1.1 — Parallel trends not tested]
ORIGINAL:
"[original text]"
REVISED:
"[new text]"
REASON: Addresses R1.1 (parallel trends) and partially addresses Ed.1 (identification)
WORD Δ: [+200 words]
RUNNING TOTAL: [N] words ([N] remaining in budget)
─────────────────────────────────────────────
Revision writing standards:
Match the voice and tense of the surrounding text
Do not introduce new claims not in the response letter
Solve exactly the reviewer's concern — no over-revision
Preserve strong existing text; change only what is needed
R2+ rule: Minimal changes only. Do not rewrite paragraphs that were not flagged.
Step 3: Consistency Check
After all revisions:
Table and figure references in text still match the actual tables/figures
Table and figure numbering is sequential (no gaps or duplicates)
Abstract accurately reflects any changed findings or framing
Contribution statement in Introduction reflects revisions
Hypothesis labels (H1, H2...) are consistent across Theory, Results, Discussion
All hypothesis results are discussed (no orphan hypotheses)
Word count is within journal limit after all revisions
All [CITATION NEEDED] markers from revision are flagged for /scholar-citation
NEW citations introduced during R&R are verified via local library, CrossRef, Semantic Scholar, or OpenAlex — not inserted from Claude's memory alone
Claim verification (MANDATORY): All prose claims attributing findings to newly added citations are checked against Knowledge Graph or PDF text — run scholar-citation Step V-3.5. R&R citations added under time pressure are the highest risk for mischaracterization. Flag all 7 marker types: [CLAIM-REVERSED], [CLAIM-MISCHARACTERIZED], [CLAIM-OVERCAUSAL], [CLAIM-UNSUPPORTED], [CLAIM-WRONG-POPULATION], [CLAIM-IMPRECISE], [CLAIM-NOT-CHECKABLE]. Correct all error-level markers before saving.
Reference list includes all newly cited works and removes any dropped citations
Tracked changes / blue text marking is applied consistently
No new typos or grammatical errors introduced by revisions
Run the claim verification gate on the revised manuscript:
Step 3a: New-Analysis Gate (MANDATORY when reviewers request new analyses)
Purpose: R&R reviewers routinely ask for new regressions (add state FE, cluster SEs differently, subset sample, different outcome). The highest failure mode in R&R is running the new analysis without the rigor of the original pipeline, then dropping numbers into the response letter via prose paraphrase. This gate forces new analyses through the same contract as the initial study.
Trigger: any "Items Requiring Author Action" in the response strategy that involves re-running or adding a regression / subset / specification.
Protocol — every new analysis dispatched from an R&R response must:
Generate the analysis script under ${PROJ}/scripts/rr-NN-[description].R, do NOT execute yet.
Run scholar-code-review in statistics + data-handling + correctness mode against the new script, using the Phase 3 design blueprint (if available) or the reviewer's specification as compliance reference. Apply the Code-Review Fix Loop from cat "${SCHOLAR_SKILL_DIR:-.}/.claude/skills/_shared/code-review-fix-loop.md". CRITICAL halts.
The new script must emit ${PROJ}/tables/rr-results-registry.csv and (if hypothesis-bearing) ${PROJ}/tables/rr-adjudication-log.csv in the same schemas as the originals, appended or separate.
Execute in a clean R session, then run plausibility + direction-consistency + (for ASR/AJS/Demography/Nature/Science) clean-room re-run checks from cat "${SCHOLAR_SKILL_DIR:-.}/.claude/skills/_shared/phase-runtime-sanity.md". CRITICAL halts.
Disk-citation discipline in the response letter (Step 4 below): every numeric claim from a new analysis must carry [rr-results-registry.csv row=X model_id=Y] — never a prose paraphrase of the agent's return text.
Skip only if: the reviewer's comment is handled without new statistics (rewording, adding citations, arguing against the request, noting a limitation).
Step 3b: Verification Gate (scholar-verify)
Purpose: R&R revisions often introduce new numbers, change table references, or alter statistical claims. Run scholar-verify on the revised manuscript to catch inconsistencies introduced during revision. When new analyses were run in Step 3a, verify-numerics and verify-logic MUST also compare prose claims against rr-results-registry.csv and rr-adjudication-log.csv.
New analyses requested by reviewers may have been transcribed incorrectly
Revised text may reference old (pre-revision) numbers
New tables/figures added during revision need completeness verification
If no raw outputs exist: The manual consistency checklist in Step 3 above is sufficient. Proceed to Step 4.
Gate decision:
0 CRITICAL issues: Proceed to Step 4.
1+ CRITICAL issues: Fix before proceeding. Include fixes in the response letter as additional revisions made during consistency checking.
Add any verification-driven fixes to the revision tracking in Step 4's Revision Summary under a "Post-revision verification fixes" subsection.
Step 3c (Optional): External Review via Codex (scholar-openai)
Purpose: For R&R revisions that involved substantial new analysis or major rewriting, run an independent external review via OpenAI Codex agents for a second opinion.
Trigger: User opts in, OR Step 3b found CRITICAL issues (cross-validate with external model).
cat .claude/skills/scholar-openai/SKILL.md
Run scholar-openai in the appropriate mode:
code — if R&R involved new analysis scripts
stats — if R&R involved new tables/numbers
full — if R&R was comprehensive
Cross-reference Codex findings with Step 3b findings. Issues confirmed by both get highest confidence.
Step 4: Produce Revision Summary
===== REVISION SUMMARY =====
Round: [R1 / R2 / R3]
Total revisions: [N] | Critical: [N] | Major: [N] | Minor: [N]
Changes by section:
Abstract: [description]
Introduction: [description]
Theory: [description]
Methods: [description]
Results: [description]
Discussion: [description]
Tables/Figures: [description]
Appendix: [description]
Word count: Original [N] → Revised [N] (limit: [N])
Consistency check: [PASSED / FAILED — list failures]
Items requiring author attention (beyond what was revised here):
• [Any items beyond what was revised — e.g., new analysis to run, data to check]
• [e.g., "Run new event study specification — requires re-running regression in R"]
• [e.g., "Obtain original data for sensitivity analysis R2.3 requested"]
Recommended next steps:
1. [Run pending analyses]
2. [/scholar-citation insert — to add all new citations to reference list]
3. [/scholar-respond cover-letter — to write the R&R cover letter]
MODE 4: RESUBMISSION STRATEGY
Use after a rejection to diagnose the root cause, select a new target journal, reframe the manuscript, and write a cover letter for resubmission.
Step 1: Triage the Rejection Decision
Decision type
What it means
Your response
Desk reject
Out of scope or below threshold; no external review
Reframe for a different journal; do not resubmit to same journal without major restructuring
Reject after review
External review; editor says fatal flaws
Diagnose root cause (Step 2); major revision before next submission
Reject with invitation
Rare; signals openness if specific concerns are addressed
Treat as conditional R&R; respond to each concern; resubmit as new submission with detailed cover letter
R&R declined / expired
You declined or missed deadline
Same as reject after review
Step 2: Diagnose Root Cause
Before resubmitting anywhere, identify the failure mode:
Root cause
Diagnosis signs
Fix
Typical time
Scope mismatch
Desk reject; "outside our scope"; no substantive critique
Change framing and opening, not the paper
1–2 weeks
Contribution threshold
"Interesting but not transformative enough for [journal]"
Assess if genuine advance is possible; or move down the journal ladder
2–4 weeks
Fatal methodological flaw
"Identification assumption untenable"; "N too small for the claims"
Fix before resubmitting ANYWHERE — new reviewers will raise it too
1–3 months
Framing mismatch
Sociology paper rejected by Nature: "too specialized"; Nature paper rejected by sociology: "not theoretical enough"
Substantive reframing of introduction and contribution
Stratification / inequality:
ASR → AJS → Social Forces → Social Problems → Sociological Quarterly → Research in Social Stratification & Mobility
Demography / population:
Demography → Population and Development Review → Population Studies → Journal of Marriage and Family → Social Science Research
Race / ethnicity:
ASR → AJS → Du Bois Review → Social Forces → Ethnic and Racial Studies → Sociology of Race and Ethnicity
Gender / family:
ASR → Gender & Society → Journal of Marriage and Family → Social Forces → Journal of Family Issues
Political sociology:
ASR → APSR → AJS → Social Forces → Mobilization → Political Research Quarterly
Culture / knowledge:
AJS → ASR → Poetics → Cultural Sociology → Theory and Society
Linguistics / language & society:
Language in Society → Journal of Sociolinguistics → Language Variation and Change → Journal of Language and Social Psychology → Applied Linguistics
NCS: Results before Methods; Nature journals: brief Methods in main text + detailed in Supplementary
Journal-specific conventions
Reframed contribution paragraph template:
"[Target journal]'s readership will recognize [the puzzle or debate that motivates the paper]. Despite [what is known], [what remains unknown or contested]. We address this gap by [what we do]. Our contribution is [specific advance: new method / new population / resolved debate / boundary condition]. This matters because [why the target audience cares, in their own terms]."
Step 5: Write the Resubmission Cover Letter
Dear [Dr. LastName / Editor],
Please find enclosed our manuscript, "[Title]," for consideration for
publication in [New Journal Name].
[1–2 sentences on why this paper is a strong fit for this journal's scope
and readership — be specific about the journal, not generic.]
[1–2 sentences summarizing the key contribution and findings.]
This manuscript has been substantially revised since an earlier version
was reviewed elsewhere. The current version includes [summary of major
revisions: e.g., "a new event study analysis, an expanded theory section,
and a reframed contribution argument"]. We believe these revisions
significantly strengthen the paper.
[Optional — pre-empt a predictable concern]:
"We note that while our design is observational, we conduct [specific
robustness checks] in Appendix B that bound the potential confounding."
The manuscript is [N] words (within [journal]'s [N]-word limit), has
not been published elsewhere, and is not under review at another journal.
We look forward to your consideration.
Sincerely,
[Name, Title, Affiliation, Email]
Step 6: Lessons Learned Log
Document what went wrong and what was fixed for future reference:
===== POST-REJECTION LESSONS LEARNED =====
Previous journal: [journal]
Decision: [desk reject / reject after review]
Root cause: [from Step 2]
What reviewers said (key themes):
1. [Theme 1]
2. [Theme 2]
What we changed:
1. [Change 1 — addresses theme 1]
2. [Change 2 — addresses theme 2]
What we did NOT change (and why):
1. [Element preserved — rationale]
New target: [journal]
Key differences in framing:
1. [Difference 1]
2. [Difference 2]
Pre-emptive defenses added to manuscript:
1. [Defense 1 — anticipating concern X]
MODE 5: R&R COVER LETTER
Use to write a standalone cover letter for an R&R resubmission (separate from the point-by-point response letter).
Step 1: Gather Information
Identify:
Journal name and manuscript number
Editor name (if known)
R&R round (R1, R2, R3)
Key changes made (from Mode 3 revision summary, or user input)
Decision received (Major Revision / Minor Revision)
Step 2: Draft the Cover Letter
R1 Cover Letter (Major Revision):
Dear [Dr. LastName / Editor],
Please find enclosed our revised manuscript, "[Title]" (Manuscript #:
[xxx]), submitted in response to the [Major/Minor] Revision decision of
[date].
We are grateful to the Editor and [N] reviewers for their constructive
and detailed feedback. We have carefully revised the manuscript to
address all concerns raised. The major changes include:
1. [Major change 1 — briefly, 1 sentence]
2. [Major change 2]
3. [Major change 3]
[Optional: 1–2 sentences addressing the editor's specific priority if
one was identified in the decision letter.]
A detailed, point-by-point response to each reviewer comment is enclosed
separately. All changes in the manuscript are marked in [blue text /
tracked changes].
The revised manuscript is [N] words, within the journal's word limit.
We believe the revisions significantly strengthen the paper and address
all reviewer concerns.
Thank you for the opportunity to revise. We look forward to your
decision.
Sincerely,
[Name, Title, Affiliation, Email]
R2 Cover Letter (second revision):
Dear [Dr. LastName / Editor],
Please find enclosed our second revision of "[Title]" (Manuscript #:
[xxx]).
We thank the Editor and reviewers for their continued engagement with
our work. In this revision, we have addressed all [N] remaining points:
1. [Change 1]
2. [Change 2]
The changes in this round are limited to the specific concerns raised.
A point-by-point response is enclosed.
The manuscript is [N] words. We hope the revised version is now suitable
for publication in [Journal].
Sincerely,
[Name, Title, Affiliation, Email]
R1 Cover Letter (Minor Revision):
Dear [Dr. LastName / Editor],
Please find enclosed our revised manuscript, "[Title]" (Manuscript #:
[xxx]), addressing the Minor Revision points from [date].
We have made all requested changes, which are detailed in the enclosed
response letter. The key revisions are:
1. [Change 1]
2. [Change 2]
We believe the manuscript is now ready for publication in [Journal].
Sincerely,
[Name, Title, Affiliation, Email]
Full Pipeline: Simulate → Respond → Revise
When invoked without a specific mode, or with all:
1. Read the manuscript
2. SIMULATE: Run 3–4 journal-calibrated reviewer agents in parallel
→ produce severity matrix + revision roadmap
3. Ask: "These are your simulated reviews. Would you like to:
(a) Draft a response letter treating these as real reviews
(b) Use these to identify weaknesses before submitting
(c) Both"
4. RESPOND: Produce triage dashboard → draft response letter
5. REVISE: Build word-budget revision plan → execute via /scholar-write revise
6. Final check: word count, consistency, formatting
7. COVER-LETTER: Draft the R&R cover letter
Verification Subagent
After completing any mode, run a verification check via Task tool:
"You are verifying a scholar-respond output. Check the following:
For MODE 1 (simulate): (1) Each reviewer addresses journal-specific concerns from the calibration table; (2) reviews are realistic in length and tone; (3) severity matrix is consistent with review content; (4) revision roadmap addresses all critical and major items; (5) no reviewer concern is missing from the action plan.
For MODE 2 (respond): (1) Every reviewer comment has a numbered response — no skipped items; (2) every response includes specific revision text or explains why none was made; (3) cross-reviewer overlaps are noted; (4) conflicting demands are resolved with rationale; (5) Changes Summary Table matches the actual changes described; (6) word count is tracked and within limits; (7) all reference library/CrossRef lookups are documented.
For MODE 3 (revise): (1) All revision plan items are executed; (2) diffs are provided for each change; (3) word budget is tracked; (4) consistency check passes; (5) no over-revision beyond what reviewers asked.
For MODE 4 (resubmit): (1) Root cause diagnosis is specific and actionable; (2) journal selection is justified; (3) reframed introduction matches new journal's conventions; (4) cover letter is journal-specific, not generic.
For MODE 5 (cover-letter): (1) Tone matches the R&R round; (2) key changes are listed concisely; (3) manuscript number and editor name included if available; (4) word count stated.
Flag any issues found. Output: [pass/fail] + [list of issues if any]."
Save Output
After completing any mode, save files using the Write tool.
Version collision avoidance (MANDATORY — RUN BEFORE EVERY Write tool call): Read and follow the version collision avoidance protocol in .claude/skills/_shared/version-check.md. You MUST run the version-check Bash block to determine the correct save path BEFORE calling the Write tool. The Bash block prints SAVE_PATH=... — use that exact printed path in the Write tool call. Do NOT hardcode a path from the filename template. Shell variables do NOT persist between Bash calls, so re-derive $BASE in every new Bash call. NEVER overwrite an existing file.
File 1 — Response Log (Internal Record)
Purpose: Internal record of strategy decisions. Not for submission.
Contains the full formatted response letter (Mode 2), R&R cover letter (Mode 5), or resubmission cover letter (Mode 4) with all reviewer comments, responses, and revision descriptions filled in.
Contains the full revision plan with word budgets, diffs, and consistency check results.
Confirm all saved file paths to the user.
Convert Response Letter to Submission Formats (MANDATORY for Modes 2, 4, 5)
Journals require the response letter as .docx or .pdf, not .md. After writing File 2 (the submission-ready letter) via the Write tool, convert it with pandoc. Shell variables do NOT persist across Bash tool calls, so derive BASE from the exact path used in the preceding Write call — do not attempt to re-derive from SLUG / OUTPUT_ROOT (those are not set in this Bash block).
set -euo pipefail
# CRITICAL: Replace [saved-md-path] with the EXACT path you used in the Write tool call# for File 2 (the submission-ready letter). This is the SAVE_PATH version-check.sh printed.
MD_FILE="[saved-md-path]"if [ ! -f "$MD_FILE" ]; thenecho"FAIL: response-letter .md not found at $MD_FILE — re-check the File 2 save path." >&2
exit 1
fi
BASE="${MD_FILE%.md}"echo"Converting: ${BASE}.md -> .docx, .pdf"
pandoc "${BASE}.md" -o "${BASE}.docx" \
--reference-doc="$HOME/.pandoc/reference.docx" 2>/dev/null \
|| pandoc "${BASE}.md" -o "${BASE}.docx"
pandoc "${BASE}.md" -o "${BASE}.pdf" --pdf-engine=xelatex 2>/dev/null \
|| pandoc "${BASE}.md" -o "${BASE}.pdf"ls -la "${BASE}".docx "${BASE}".pdf 2>/dev/null
Notes:
Mode 3 (revised manuscript) is exported via the scholar-write pandoc block — scholar-respond does NOT re-convert the manuscript.
If xelatex is missing, pandoc falls back to the default PDF engine; the letter is short so this is acceptable.
Verify both files exist and spot-check the .docx opens cleanly before submission.
Knowledge Graph Write-Back (post-save)
SKILL_DIR="${SCHOLAR_SKILL_DIR:-.}/.claude/skills"
KG_REF="$SKILL_DIR/_shared/knowledge-graph-search.md"if [ -f "$KG_REF" ]; theneval"$(cat "$KG_REF" | sed -n '/^```bash/,/^```/p' | sed '1d;$d')" 2>/dev/null
if kg_available 2>/dev/null; thenecho""echo"═══ Knowledge Graph ═══"echo"Reviewers may have suggested references not in your knowledge graph. Ingest them:"echo" /scholar-knowledge ingest from doi [DOI] (for each new reference from reviewers)"fifi
Close Process Log:
Run the following to finalize the process log:
SKILL_NAME="scholar-respond"
LOG_DATE=$(date +%Y-%m-%d)
LOG_FILE="${OUTPUT_ROOT}/logs/process-log-${SKILL_NAME}-${LOG_DATE}.md"if [ ! -f "$LOG_FILE" ]; then
LOG_FILE=$(ls -t "${OUTPUT_ROOT}"/logs/process-log-${SKILL_NAME}-${LOG_DATE}*.md 2>/dev/null | head -1)
ficat >> "$LOG_FILE" << LOGFOOTER
## Output Files
[list each output file path as a bullet]
## Summary
- **Steps completed**: [N completed]/[N total]
- **Files produced**: [count]
- **Errors**: [count, or 0]
- **Time finished**: $(date +%H:%M:%S)
LOGFOOTERecho"Process log saved to $LOG_FILE"
Quality Checklist
Response Letter Quality
Every reviewer comment has a numbered response — no skipped items
Every response includes a specific revision action or explains why none was made
Exact new text is quoted in the response (not just "we revised this")
Page/line numbers given for all revisions in the response letter
Response tone is respectful throughout, even when disagreeing
Changes Summary Table included and matches actual changes
Cross-Reviewer Handling
Cross-reviewer overlaps identified and noted explicitly in letter
Conflicting reviewer demands named and resolved with rationale
Editor priorities addressed first and most thoroughly
Citation and Reference Integrity
Local reference library checked for all reviewer-requested citations
CrossRef API used as fallback for citations not in local library
All newly cited works added to reference list
No orphaned citations (cited in text but missing from references)
Word Count and Formatting
Word count verified against journal limit after revisions
Word impact tracked per revision item
Abstract updated if findings or framing changed
Tracked changes / blue text applied consistently
Causal Language
Causal language audit passed: response letter and revised text maintain the manuscript's language precision — if study is non-causal, do not upgrade to causal language even when responding to reviewers. See scholar-write SKILL.md for full rule
Consistency
Table/figure references match actual tables/figures
Hypothesis labels (H1, H2...) consistent across all sections
Contribution statement in Introduction reflects revisions
No new analyses or content added beyond what reviewers requested (R2+ rule)
Round-Specific
R&R round noted in all output files
Tone calibrated to round (R1 thorough; R2 direct; R3 surgical)
R2+ new comments flagged as [NEW-IN-R2+]
Cover letter matches round and revision scope
See references/response-templates.md for complete letter templates and phrase bank.
See references/common-concerns.md for common reviewer concerns by journal and method type, conflict resolution templates, reviewer personality guide, and resubmission strategy.
