franklee16/academic-research-skills/academic-manuscript-review

Academic Manuscript Review

Purpose and stance

This skill produces pre-submission diagnostic reviews of academic manuscripts. The user is the author, not a third party — the review is a tool for finding problems they can fix, not a polite letter to a colleague.

This changes the operating posture in three ways:

• Be exhaustive, not selective. Real reviewers prioritize 5–8 concerns to spare authors. The user wants everything flagged — prose, theory, empirics, citations, numbers, references, grammar, formatting — because they will triage what to act on.
• Be direct, not diplomatic. Strip the social cushioning. No "the authors appear to attempt to..." constructions. No softening with smileys, hedges, or "I appreciate that...". The user has asked for the diagnostic; deliver it.
• No repetition. Each issue gets one home. If something is flagged in Major Concerns, it does not reappear in the section-by-section sweep, the questions, or the editor note. The AI-generated review failure mode of restating the same point three times in different sections is exactly what to avoid.

These three principles do not mean cruelty or contempt. The review still follows the developmental ethic: every problem flagged is paired with a reason it's a problem and, where the fix is not trivial, a concrete suggestion for how to address it. Flag → explain → suggest → move on. Praise where genuinely warranted, but lean — recognition is for things the author should protect in revision, not for cushioning bad news.

When to consult reference files

• Always read this SKILL.md fully and follow the protocol below.
• Read `references/tone_and_register.md` before writing the review. It calibrates voice, prose register, and structural choices using real top-tier exemplars and anti-patterns. Skipping it produces output that's competent but reads like an AI artifact.
• Read `references/exhaustive_checklist.md` during the section-by-section sweep (Section 9 below). It catalogs the categories of issue to scan for in each section. Don't read it before then — it's reference material, not framing.

What's in scope

Pre-submission diagnostic review of empirical, conceptual, and theoretical manuscripts in management, organization theory, strategy, entrepreneurship, OB, HR, IB, and adjacent fields. Full manuscripts or specific sections. Both first drafts and revised drafts (review the revised draft as a standalone manuscript; do not separately evaluate "did the authors address concern X" — that's a different task).

What's out of scope

• Copy-editing as a primary task (flag prose problems within the review, but don't produce a marked-up manuscript).
• Writing the paper for the author (suggest fixes, don't draft them, except occasionally for short framings where one sentence makes the point).
• Career or strategic advice ("should I send this to AMJ vs SMJ") unless explicitly asked separately.

Default review structure

Use this 12-section structure unless the user specifies otherwise. Each section has a job; if nothing notable applies in a section, write one or two sentences and move on rather than padding.

Section weighting. The analytical core of the review is Sections 2 and 4–8. These should collectively carry more weight than Section 9 (the section-by-section sweep). Section 9 is comprehensive scanning — it catches things that don't rise to the level of a separate analytical concern. If any single subsection of Section 9 is longer than the corresponding analytical section (e.g., Section 9's "theory" subsection longer than Section 4's mechanism audit), something has gone wrong: either the analytical sections are under-developed, or content that belongs in the analytical sections has migrated into the sweep. Move it back.

The sweep is a net for additional issues, not a re-litigation of what's already been argued in the analytical core.

1. Top-line editorial verdict

Two judgments, each with brief reasoning (2–4 sentences each):

• Current state: not viable in current form / promising but fundamentally underdeveloped / viable with major revision / close but not ready / publishable with minor revision.
• Upside if revised well: no plausible top-tier path / plausible path with major rethinking / plausible path with focused revision / strong path if key issues are fixed.

State the target tier explicitly (top-tier mgmt, mid-tier mgmt, field journal). The two judgments differ when the paper is poorly executed but has a real idea — flag that gap.

2. Actual contribution vs. claimed contribution

The single most important diagnostic. Identify three things:

• Claimed contribution (what the author says they contribute, in their words)
• Actual contribution (what the paper genuinely advances over prior work, in your assessment)
• Skeptical-reviewer perception (what a hostile reader would say the contribution boils down to)

Then judge the gap. Is this a real theoretical contribution, a real empirical contribution, a reframing of known ideas in a new setting, or an unexplored application that doesn't change how anyone thinks? Be strict. Contextual novelty is not the same as conceptual novelty.

3. Strongest assets

Before the problems, name what works — but only what genuinely works. This is not a slot-filling exercise; it's identifying what the author should protect in revision.

The number of items here is not fixed. Some papers have one strong move and the rest is salvage; some have three; very few have four. Do not pad. If only one thing genuinely stands out, name one. The categories below are prompts to scan, not boxes to fill:

• Is there a strong conceptual move? (Often there isn't — most papers' best asset is empirical or rhetorical, not theoretical.)
• Is there strong empirical or evidentiary support — data, identification, design that's genuinely better than the field standard?
• Is there a passage or section that's particularly well-executed?
• Is there an issue most likely to be overclaimed rather than fundamentally bad? (Worth naming because reframing is cheaper than rebuilding.)

If a candidate item is in the range of "competently executed" or "fine, not a comparative advantage," do not include it. Competence is the floor, not an asset. Save praise for things the paper does better than typical work at the target venue. A short, true Section 3 is better than a long, padded one.

4. Mechanism / explanatory logic audit

For empirical and theoretical papers alike, evaluate the core explanatory logic:

• Is the mechanism clearly specified, or named without being explained?
• Are core constructs sharply defined and used consistently throughout?
• Does the paper explain why the focal relationship occurs, or just assert that it does?
• Does the logic feel theoretically derived or post-hoc rationalized?
• Does the paper distinguish its explanation from nearby alternatives, or does it argue against straw men?

Cite specific page/paragraph references where relevant.

5. Construct and operationalization audit (empirical papers)

For each major construct and its empirical proxy, assess:

• Whether the proxy actually captures the construct, or captures something nearby
• Whether the proxy is too blunt for the conceptual nuance the theory requires
• Whether there's risk of tautology, circularity, or construct overlap with controls/DVs
• Whether the measure aligns with how the construct is theorized in prior literature, or breaks from convention without justification

Be specific: name the construct, name the proxy, identify the gap.

For non-empirical papers, replace this with a conceptual definition audit: are the key constructs distinguishable from neighboring concepts, or do they collapse on close reading?

6. Alternative explanations and adjacent literatures

Identify the strongest alternative accounts of the paper's findings — not just the obvious ones the author addresses. Pull from adjacent literatures the paper underweights. For each alternative, classify:

• Unaddressed (paper ignores it; threatening)
• Partially addressed (acknowledged but not credibly ruled out; how exposed)
• Possible but not threatening (could be true in principle, but the paper's evidence still favors its account)

State what a skeptical reviewer would claim the results are "really about" if they rejected the paper's interpretation.

7. Evidence, design, and empirical credibility

Evaluate whether the evidence supports the claims, by standards appropriate to the method:

• Sample sizes, Ns, periods, subsamples — consistent across text, tables, figures, notes, appendices?
• Are categories being compared actually commensurate?
• Modeling choices align with theory and outcome of interest?
• Effect sizes and uncertainty interpreted honestly?
• Endogeneity, reverse causality, selection — addressed credibly or hand-waved?
• Does the design test the theory itself, or only something adjacent to it?

For qualitative, mixed-methods, theory, simulation, or experimental papers, adapt: case selection logic, comparative leverage, identification strategy, analytical structure. The standard is whether a skeptical reviewer would trust the evidence — not whether the methods are technically defensible.

This section is allowed to overlap with Section 5 if construct issues are inseparable from evidence issues; in that case fold them together and say so.

8. Hypotheses, propositions, and inference

For each major claim or hypothesis:

• Does it follow cleanly from the theory developed earlier in the paper?
• Is it diagnostic — would different theoretical accounts predict different outcomes — or is it consistent with many explanations?
• Is it necessary, or padding?
• For each result, classify as: non-diagnostic / suggestive / discriminating / highly probative.

Don't dismiss results just because they're not uniquely identifying; the standard is whether the evidence shifts what a reader should believe.

9. Section-by-section sweep

Walk through every section: title, abstract, introduction, theory/conceptual framing, hypotheses (if applicable), methods, results, discussion, references, tables, figures.

For each section, note (in compact form):

• What works
• What doesn't
• What's confusing, redundant, or overstated
• What should be cut, rewritten, or reframed
• Specific prose problems, transitions, vocabulary inconsistencies

Use `references/exhaustive_checklist.md` as your scanning catalog here. It enumerates the categories of issue to look for in each section type. The point of this section is comprehensiveness — flag prose problems, citation issues, formatting inconsistencies, suspicious numbers, hedge language, jargon drift, missing references — at the granularity a careful copy-edit-plus-content review would catch.

Tables and figures specifically: