Instruction file imported from nachtflug6/academic-writing-template (.github/instructions/review.instructions.md). Copyright stays with the author.

Review Instructions

Use these guidelines when reviewing the workspace before major handoff, internal review, defense preparation, or submission.

Purpose

The goal of review is not only to catch formatting problems. It is to identify weaknesses that could cause trouble during supervision, opponent review, examination, peer review, or submission.

A good review should assess:

• technical integrity,
• manuscript-evidence separation,
• argument clarity,
• scope discipline,
• claim defensibility,
• likely reviewer or opponent challenges,
• and submission readiness.

Keep the review structured. Distinguish clearly between:

• format/integrity issues,
• writing/clarity issues,
• evidence/claim issues,
• argument/content issues,
• and handoff/submission issues.

Do not silently rewrite during review unless explicitly asked. Review should primarily identify, classify, and prioritize issues.

1. Review modes

A. Format and integrity review

Check technical and workspace consistency. Focus on:

• LaTeX integrity,
• file structure,
• references,
• includes,
• package consistency,
• venue requirement alignment,
• submission-package readiness.

B. Content and argument review

Check whether the manuscript's reasoning is clear, complete enough, and defensible within its stated scope. Focus on:

• missing motivations,
• unexplained choices,
• weak transitions in argument,
• unclear contribution framing,
• missing definitions,
• unsupported reasoning jumps,
• scope inconsistencies,
• unaddressed alternatives or omissions.

C. Opponent / examiner challenge review

Read like a skeptical but fair opponent, reviewer, or examiner. Ask:

• Why was X not discussed?
• Why was Y chosen instead of Z?
• What assumptions does this depend on?
• Is this claim too broad?
• Is the limitation acknowledged clearly enough?
• Is the contribution distinct and justified?
• Would a reviewer ask for comparison, citation, or clarification here?
• Is one case being stretched into a general claim?
• Is this statement vulnerable to an obvious counterexample?
• Where would an opponent push on novelty, rigor, generality, or interpretation?

D. Evidence and defensibility review

Check whether claims are supported at the right strength. Focus on:

• traceability from claims to evidence,
• distinction between evidence and interpretation,
• bounded wording,
• preservation of uncertainty,
• causal precision,
• novelty bounding,
• citation gaps,
• overclaim risk,
• mismatch between wording and support.

E. Delivery and handoff review

Check whether the workspace is ready for the next person or next phase. Focus on:

• current task state,
• known open issues,
• ambiguous files,
• packaging readiness,
• visible next actions,
• unresolved risk items.

2. Scope checks

Confirm that:

• manuscript and evidence boundaries are respected,
• bibliography files remain in refs/,
• cite keys are unchanged,
• standards constraints in standards/requirements.md are considered (formal/template rules in standards/formatting-requirements.md, benchmark baseline in standards/benchmark-requirements.md),
• files are located and named consistently,
• ambiguous materials are explicitly identified rather than silently reclassified.

3. LaTeX integrity checks

Check that:

• no \label{}, \ref{}, or \cite{} commands are broken,
• include/input paths remain valid,
• figure/table references still resolve correctly,
• environments remain intact,
• no accidental package additions were made unless requested,
• section structure remains coherent after edits,
• comments and important annotations were not removed.

4. Manuscript editing compliance checks

Verify that edits to manuscript files followed the guidelines in .github/instructions/manuscript.instructions.md.

Check that:

• minimum intervention was applied: changes are targeted rather than wholesale rewrites,
• the author's scientific intent, results, and argument structure are unchanged,
• no new theoretical content or claims were introduced,
• no sections were restructured without explicit instruction,
• the author's voice has not been distorted,
• equations, notation, symbols, and formal derivations were not altered,
• author comments and annotations were not removed,
• bibliography metadata was not changed,
• edits favor clarity and flow improvements: sentence simplification, stronger topic sentences, better transitions, sharper paragraph unity,
• epistemic status is visible: distinctions between observation, inference, interpretation, and hypothesis are preserved or clarified,
• claim scope and certainty are aligned with the evidence described,
• language-risk edits softened or flagged overclaims, absolutes, causal overreach, novelty inflation, and vague evaluative language.

Ask:

• Were edits minimal and targeted?
• Was anything added that was not already supported by the source material?
• Were any conditions, limitations, or distinctions erased?
• Does the edited text satisfy the strength-mismatch principle: is the wording no stronger than the evidence, and is the scope no broader than what was shown?

5. Writing and readability checks

Review whether the manuscript is readable and scientifically well-formed.

Check for:

• unclear or overloaded sentences,
• weak topic sentences,
• abrupt paragraph transitions,
• unclear referents,
• excessive abstraction,
• unnecessary jargon density,
• vague evaluative language,
• poor paragraph unity,
• missing explanation of why one idea follows another,
• formal but cognitively difficult prose.

Also check whether the prose remains:

• precise without being opaque,
• rigorous without being stiff,
• concrete without becoming informal,
• accessible without losing substance.

6. Content and argument review checks

Review the manuscript's reasoning, not just its wording.

Check for:

• unclear problem framing,
• weak statement of contribution,
• unclear research gap,
• unexplained design or method choices,
• missing rationale for important decisions,
• unsupported argumentative jumps,
• concepts introduced without definition,
• distinctions that should be explicit but are blurred,
• discussion sections that summarize without interpreting,
• limitations that are missing or too weakly stated,
• conclusions that extend beyond the actual evidence.

Typical review questions include:

• Why is this problem important in this specific scope?
• Why was this method, dataset, case, or comparison chosen?
• Why are some related approaches not discussed?
• What would a critical reader say is missing?
• What alternative interpretation exists?
• Does the manuscript answer the questions it raises?

7. Opponent / examiner challenge checks

Read parts of the manuscript as if preparing challenges for a defense or peer review.

Look for places where an opponent could reasonably ask:

• Why is X excluded?
• Why is Y included?
• Why this framing and not another?
• Why is this comparison sufficient?
• Why should this result generalize?
• What exactly is new here?
• What is the evidence for that wording?
• Is this correlation being presented as causation?
• Is the novelty claim bounded enough?
• Is this claim true only in this setting?
• What would happen if a contrary example were raised?

Flag sentences or sections that are easy to challenge because they are:

• too broad,
• too absolute,
• under-cited,
• poorly bounded,
• weakly motivated,
• or unclear about assumptions.

8. Evidence integrity checks

Check that:

• claims remain traceable to evidence notes,
• uncertainty is marked rather than erased,
• no invented data, figures, or outcomes appear,
• evidence files do not get silently turned into manuscript claims,
• limitations, null results, and contradictory evidence are not suppressed,
• claim wording matches the strength of support,
• cited-background claims are not presented as if directly established by the current work.

9. Language-risk checks

Review for wording likely to invite reviewer or examiner attack.

Flag:

• absolute formulations,
• overclaims,
• causal overreach,
• novelty inflation,
• vague quantifiers,
• promotional wording,
• rhetorical dismissal of alternatives,
• scope creep from one case to many.

Ask:

• Is the wording stronger than the evidence?
• Is the scope broader than what was shown?
• Would one counterexample damage the sentence badly?
• Would this survive a skeptical "how do you know?" question?

10. Delivery checks

Confirm that:

• tasks/status.md reflects the current state,
• tasks/backlog.md reflects next actions,
• submission/package-checklist.md reflects packaging readiness,
• unresolved issues are visible,
• ambiguous files are explicitly listed rather than force-classified,
• known risks are documented rather than left implicit.

11. Review output format

When producing a review, structure it as:

1. Overall verdict

   • Ready / Nearly ready / Needs revision / High risk

2. Top issues first

   • most important problems affecting defensibility, clarity, integrity, or submission readiness

3. Issues by category

   • format/integrity
   • manuscript editing compliance
   • writing/readability
   • content/argument
   • opponent/examiner vulnerability
   • evidence/claim support
   • delivery/submission

4. Questions a reviewer or opponent is likely to ask

   • list the strongest likely challenges

5. Priority fixes

   • what should be fixed first before handoff or submission

Keep the review diagnostic and prioritized. Do not drown major risks in minor cosmetic comments.

12. Standard for success

A successful review should make it clear:

• whether the workspace is technically sound,
• whether manuscript edits followed the guidelines in manuscript.instructions.md,
• whether the manuscript is readable and coherent,
• whether the argument is defensible,
• where an examiner or reviewer is most likely to attack,
• whether claims are properly supported and bounded,
• and what must be fixed before the next handoff or submission.

13. Operational review templates

Use these templates for consistent review outputs.

A. Whole-thesis structure/readiness template

1. Verdict: Strong / Adequate / Weak / At Risk.
2. Dense summary.
3. Top strengths (5).
4. Top weaknesses (5).
5. Missing core vs optional elements.
6. Top 10 fixes first (issue | why | first fix).

B. Blockers and priority plan template

1. Blockers by category (research | writing | structure | compliance | technical | external) with impact and owner.
2. Dependency-aware sequencing.
3. Next 7 concrete actions.
4. Time estimates.
5. Escalations/questions for advisor or co-authors.

C. Feedback integration template

1. Feedback matrix (chapter | issue | priority | owner).
2. Integration order with rationale.
3. Conflicts and clarification questions.
4. Cross-chapter consistency risks.
5. Verification checklist after implementation.

D. Citation/reference audit template

1. Missing/unmatched cite keys.
2. Under-cited hotspots by chapter.
3. Bibliography data-quality issues.
4. Orphaned references.
5. Priority fixes.

E. Compliance/figures/format template

1. Requirement-by-requirement compliance status (pass/fail + evidence note).
2. Figure/table findings.
3. Formatting/navigation risks.
4. Submission blockers.
5. Final preflight checklist.

F. Language-risk template

1. Overall language-risk verdict.
2. Dense risk-profile summary.
3. Repeated risky patterns and safer wording direction.
4. Category findings (absolutes, overclaims, causal overreach, novelty inflation, vague quantifiers/evaluation, rhetorical tone, first-person risk).
5. Top language fixes first (pattern | why risky | first wording move).

G. Chapter benchmark template

Use this chapter-level structure when reviewing a single chapter:

1. Verdict: Meets / Nearly meets / Below expectations.
2. Dense summary.
3. Top strengths (5).
4. Top weaknesses (5).
5. Missing core vs optional chapter elements.
6. Boundary issues.
7. Reviewer-risk issues.
8. Rubric notes by chapter role criteria.
9. Top 10 fixes first (issue | why | first fix).

Chapter-specific expectations and rubric anchors are defined in standards/benchmark-requirements.md.

14. Constraints during review execution

When operating in review mode:

• do not rewrite manuscript sections unless explicitly asked,
• do not invent evidence, citations, or metadata,
• do not replace high-impact risks with cosmetic comments,
• do not claim full compliance without requirement-level evidence,
• prioritize defensibility and submission risk over style preference.