review-paper
Skill8 starsChanged 4 months agoMITView on GitHub
Manuscript review covering argument structure, econometric specification, citation completeness, likely referee objections. Use when getting substantive academic feedback on a paper, not surface edits.
10 other projects use a version of this, 3 of them edited.
Adopted by
Most recently taken up by
- pedrohcgs/claude-code-my-workflowedited, first seen Nov 7, 2026
- MatthewMair42/claude-configtheir own version, first seen Oct 12, 2026
- ericluo04/claude-academic-workflowtheir own version, first seen Oct 9, 2026
- zhanyinx/skillstheir own version, first seen Sep 3, 2026
- claesbackman/AI-research-feedbacktheir own version, first seen Aug 26, 2026
- ExplorerFreda/fskillstheir own version, first seen Aug 5, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 3
- pedrohcgs/claude-code-my-workflow★ 1.7k14% same
- maxwell2732/paper-replicate-agent-demo★ 13476% same
- maxwell2732/my-submission-formatting-agent★ 10648% same
Wrote their own version · 7
- claesbackman/AI-research-feedback★ 523same name
- ericluo04/claude-academic-workflow★ 26same name
- ExplorerFreda/fskills★ 11same name
- Max-astro/paper-reading-skills★ 9same name
- luischanci/claude-code-research-starter★ 1same name
- zhanyinx/skills★ 1same name
- MatthewMair42/claude-config★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- data-analysisSkill
- data-finderSkill
- lit-reviewSkill
- new-projectSkill
- proofreadSkill
- quality-gateSkill
- research-setupSkill
- review-rSkill
- revise-paperSkill
- session-logSkill
- validate-bibSkill
- write-paperSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Locate and read the manuscript: Check:
- Distinctive: line 103
Wait for all three to complete.
- Distinctive: line 257
Be constructive: Every criticism should come with a suggestion.
- Distinctive: line 260
Distinguish fatal flaws from minor issues: Not everything is equally important.
- Distinctive: line 262
Do NOT fabricate details: If you can't read a section clearly, say so.
“Common” is measured: kept by at least 6 in 10 of the 10 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.
Files
Manuscript Review
Produce a thorough, constructive review of an academic manuscript — the kind of report a top-journal referee would write.
Input: $ARGUMENTS — path to a paper (.tex, .pdf, or .qmd), or a filename in manuscripts/ or references/papers/.
Steps
- Locate and read the manuscript. Check:
- Direct path from
$ARGUMENTS manuscripts/$ARGUMENTSreferences/papers/$ARGUMENTS- Glob for partial matches in
manuscripts/andreferences/papers/
- Direct path from
- Read the full paper end-to-end. For long PDFs, read in chunks (5 pages at a time).
- Read
references/domain-profile.mdfor field and top journals — use these to calibrate the referee perspective (see Principles). - Dispatch three reviewer agents in parallel via Task (one message, three Task calls — see below):
domain-reviewer— substantive correctness through 6 lenses (including design-specific diagnostic reporting)adversarial-reviewer— hostile-referee attack on the paperfresh-eyes-reviewer— first-time reader perspective
- Evaluate writing quality and presentation (dimensions 5-6) — the skill handles these directly while the agents run.
- After all agents complete, merge findings into the unified report:
fresh-eyes-revieweroutput → "Fresh Eyes Read" sectiondomain-revieweroutput → "Major Concerns" and "Minor Concerns" sectionsadversarial-revieweroutput → primary source for "Referee Objections" and any FATAL-severity entries in "Major Concerns"- Skill's own Dim 5-6 evaluation → "Writing Quality" and "Presentation" entries in concerns
- Produce the unified review report.
- Save to
quality_reports/paper_review_[sanitized_name].md
Step 4: Dispatch Three Reviewer Agents in Parallel
Send one message with three Task calls so the agents run concurrently. Each agent has a distinct job and they should not see each other's output.
Task 1: domain-reviewer
Task prompt: "You are the domain-reviewer agent. Review the manuscript at [path].
Research question: [from spec if available].
Apply all 6 review lenses:
1. Assumption stress test
2. Derivation verification
3. Citation fidelity
4. Code-theory alignment
5. Backward logic check
6. Design-specific diagnostics audit — verify that the paper reports the actual numerical results of the diagnostics the claimed design demands (parallel-trends event-study coefficients, McCrary t-stat, Olea-Pflueger first-stage F, AR weak-IV CI, in-space placebo permutation, balance table, etc., depending on design). Treat a missing required diagnostic as CRITICAL.
Also check cross-document consistency.
Follow the domain-reviewer agent instructions and return your full substance review report."Task 2: adversarial-reviewer
Task prompt: "You are the adversarial-reviewer agent. Your job is to attack the
paper at [path] — find the strongest possible critique a hostile referee would make.
Research question: [from spec if available].
Target venue: [from domain-profile.md if set, else 'unspecified — use top field journal bar'].
Apply all 5 attack lenses:
1. Fatal flaw hunt
2. Over-claim detection
3. Alternative explanations (generate the 3 most plausible)
4. Identification weakest link
5. Rejection letter (two-paragraph desk-editor rejection)
Follow the adversarial-reviewer agent instructions and return your full adversarial report."Task 3: fresh-eyes-reviewer
Task prompt: "You are the fresh-eyes-reviewer agent. Read the paper at [path]
as a first-time reader with NO prior context. Do NOT read the project spec,
lit review, ideation file, or analysis scripts — read only the manuscript.
Execute all 5 passes:
1. Title + abstract only
2. Introduction only
3. Main result table/figure standalone
4. Full paper read
5. What stays with you
Follow the fresh-eyes-reviewer agent instructions and return your full fresh-eyes report."Wait for all three to complete. Collect their outputs for the merge step.
Steps 5-6: Evaluate Writing/Presentation, Then Merge
While the agents run, evaluate dimensions 5-6 directly (writing quality, presentation). Once all three agents return:
- Fresh Eyes Read section — paste the fresh-eyes summary (Passes 1-5 condensed + top 3 clarity issues).
- Major Concerns — combine
domain-reviewerMAJOR/CRITICAL findings from all 6 lenses +adversarial-reviewerfatal flaws + your Dim 5-6 major findings. Surface Lens 6 CRITICAL items (missing required diagnostics) at the top of Major Concerns — these are the items most likely to draw a referee rejection. - Minor Concerns —
domain-reviewerMINOR findings + your Dim 5-6 minor findings. - Referee Objections — primarily from
adversarial-reviewer: fatal flaw, top over-claims, top alternative explanations, identification weakest link. - Rejection Letter Preview — paste the adversarial-reviewer's two-paragraph rejection verbatim under a "What a desk editor might say" subheading.
Review Dimensions
1. Argument Structure
- Is the research question clearly stated?
- Does the introduction motivate the question effectively?
- Is the logical flow sound (question → method → results → conclusion)?
- Are the conclusions supported by the evidence?
- Are limitations acknowledged?
2. Identification Strategy
- Is the causal claim credible?
- What are the key identifying assumptions? Are they stated explicitly?
- Are there threats to identification (omitted variables, reverse causality, measurement error)?
- Are robustness checks adequate?
- Is the estimator appropriate for the research design?
3. Econometric Specification
- Correct standard errors (clustered? robust? bootstrap?)?
- Appropriate functional form?
- Sample selection issues?
- Multiple testing concerns?
- Are point estimates economically meaningful (not just statistically significant)?
4. Literature Positioning
- Are the key papers cited?
- Is prior work characterized accurately?
- Is the contribution clearly differentiated from existing work?
- Any missing citations that a referee would flag?
5. Writing Quality
- Clarity and concision
- Academic tone
- Consistent notation throughout
- Abstract effectively summarizes the paper
- Tables and figures are self-contained (clear labels, notes, sources)
6. Presentation
- Are tables and figures well-designed?
- Is notation consistent throughout?
- Are there any typos, grammatical errors, or formatting issues?
- Is the paper the right length for the contribution?
Output Format
# Manuscript Review: [Paper Title]
**Date:** [YYYY-MM-DD]
**Reviewer:** review-paper skill (agents: domain-reviewer, adversarial-reviewer, fresh-eyes-reviewer)
**File:** [path to manuscript]
## Summary Assessment
**Overall recommendation:** [Strong Accept / Accept / Revise & Resubmit / Reject]
[2-3 paragraph summary: main contribution, strengths, and key concerns]
## Fresh Eyes Read
*How the paper lands on a first-time reader (from fresh-eyes-reviewer agent).*
**After title + abstract — one-sentence takeaway:** ...
**After intro — what reader expects:** ...
**Main display standalone?** [Yes / Partially / No — what's missing]
**What stays with the reader:** ...
### Top 3 Clarity Issues
1. [issue with location]
2. ...
3. ...
### Top 3 Things That Landed
1. ...
2. ...
3. ...
## Strengths
1. [Strength 1]
2. [Strength 2]
3. [Strength 3]
## Major Concerns
### MC1: [Title]
- **Source:** [domain-reviewer / adversarial-reviewer / skill (writing or presentation)]
- **Dimension:** [Identification / Econometrics / Argument / Literature / Writing / Presentation / Fatal Flaw]
- **Issue:** [Specific description]
- **Suggestion:** [How to address it]
- **Location:** [Section/page/table if applicable]
[Repeat for each major concern. FATAL-severity items from adversarial-reviewer go first.]
## Minor Concerns
### mc1: [Title]
- **Issue:** [Description]
- **Suggestion:** [Fix]
[Repeat]
## Referee Objections
*Primarily sourced from the adversarial-reviewer's attack lenses — these are the tough questions a hostile referee would raise.*
### RO1: [Question]
**Why it matters:** [Why this could be fatal]
**How to address it:** [Suggested response or additional analysis]
[Include: fatal flaw as RO1, top over-claims, top alternative explanations, identification weakest link. Target 4-6 objections total.]
## What a Desk Editor Might Say
*Two-paragraph rejection letter from the adversarial-reviewer. Read this before submission.*
[Paste adversarial-reviewer's rejection letter verbatim.]
## Specific Comments
[Line-by-line or section-by-section comments, if any]
## Summary Statistics
| Dimension | Rating (1-5) |
|-----------|-------------|
| Argument Structure | [N] |
| Identification | [N] |
| Econometrics | [N] |
| Literature | [N] |
| Writing | [N] |
| Presentation | [N] |
| First-read clarity | [N] |
| Adversarial robustness | [N] |
| **Overall** | **[N]** |Principles
- Be constructive. Every criticism should come with a suggestion.
- Be specific. Reference exact sections, equations, tables.
- Calibrate to the field's top journals (from
references/domain-profile.md). An AER/QJE referee weights identification above all else; an APSR/AJPS referee weights theoretical framing and measurement validity; an ASR/AJS referee weights theoretical contribution. If no domain profile is set, flag the assumption you're making. - Distinguish fatal flaws from minor issues. Not everything is equally important.
- Acknowledge what's done well. Good research deserves recognition.
- Do NOT fabricate details. If you can't read a section clearly, say so.
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
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.

