---

name: rrwrite-critique-manuscript description: Performs adversarial critique of manuscripts, outlines, literature reviews, or other academic content against journal requirements and quality standards. arguments:

• name: target_dir description: Output directory for manuscript files (e.g., manuscript/repo_v1) default: manuscript allowed-tools:

---

Academic Critique Protocol

Scope

This skill can critique multiple types of academic content:

1. Manuscript Drafts - Full or partial manuscript sections
2. Manuscript Outlines - Structure and logical flow (manuscript_plan.md)
3. Literature Reviews - Background research summaries
4. Critique Documents - Meta-critique of other critiques

Critique Style

Apply critical, demanding critique style focused on reproducibility, clarity, and rigor.

Critique Mode Selection

Automatically detect what is being critiqued:

If critiquing manuscript_plan.md (Outline):

Focus on:

• Logical flow and narrative arc
• Section ordering and dependencies
• Evidence-to-claim mapping
• Missing sections or components
• Target journal structure compliance

If critiquing {target_dir}/literature.md:

Focus on:

• Coverage completeness (foundational, related, recent)
• Citation accuracy and verifiability
• Gap analysis clarity
• Integration guidance quality
• Balance between domains/approaches

If critiquing manuscript drafts ({target_dir}/*.md):

Focus on:

• Technical accuracy and reproducibility
• Journal-specific compliance
• Citation integrity
• Data availability
• Writing quality

If critiquing another critique document:

Focus on:

• Constructiveness of feedback
• Actionability of suggestions
• Coverage of critical issues
• Balance of criticism

---

Compliance Checks (For Manuscript Drafts)

1. Journal Specifics:

   • Nature Methods: Check word count of the Abstract (max 150 words).
   • PLOS Computational Biology: Verify presence of "Data Availability Statement" and "Ethics Statement".
   • Bioinformatics: Check that the "Abstract" has structured headers.

2. Citation Integrity:

   • Scan text for citation keys (e.g., [smith2020]).
   • Verify they exist in bib_index.md or references.bib.
   • Flag any missing keys as "HALLUCINATION RISK".

3. Figure Callouts:

   • Ensure logical ordering (Figure 1 appears before Figure 2).
   • Flag any figures in the figures/ folder that are not referenced in the text.

4. Availability Section Citations:

   • Check Data and Code Availability (or similar) sections for inappropriate citations.
   • ACCEPTABLE citations: Specific tools/platforms (Zenodo DOI, Docker, GitHub, data repositories).
   • UNACCEPTABLE citations: General methodology papers (FAIR principles, reproducibility frameworks, workflow standards).
   • Rationale: Availability sections should contain factual access information, not methodology justifications.
   • Action if violated: Flag as minor issue, recommend removing general citations and keeping only tool-specific ones.

5. Methods Section Citations:

   • Check Methods sections for abstract concept citations vs. specific tool citations.
   • ACCEPTABLE citations:
     • Specific software tools actually used (e.g., LinkML for schema validation, pandas for data processing)
     • Datasets accessed (e.g., GTDB for taxonomic data, MediaDive for media formulations)
     • Published algorithms implemented (e.g., MaxPro for experimental design)
     • Computational methods applied (e.g., graph embeddings, flux balance analysis)
   • UNACCEPTABLE citations:
     • Abstract principles (FAIR data sharing [Wilkinson2016], reproducibility frameworks)
     • General best practices papers (workflow standards, documentation guidelines)
     • Related tools NOT used (e.g., citing Manubot when not using it)
     • Methodological reviews unless specific method was implemented
   • Rationale: Methods describes what was done in THIS work, not general field principles. Abstract concepts belong in Introduction or Discussion.
   • Action if violated: Flag as minor issue. Recommend moving abstract principle citations to Introduction (for motivation) or Discussion (for broader context), keeping only tool-specific citations in Methods.

6. Results Section Citations:

   • Check Results sections for explanatory/justification citations vs. observational citations.
   • ACCEPTABLE citations:
     • Papers/datasets analyzed or benchmarked (e.g., compared performance against [Smith2020])
     • Examples of findings (e.g., "identified papers including [Example2024, Example2025]")
     • Data sources processed (e.g., sequences from [GTDB2024])
     • Tools whose performance was measured
   • UNACCEPTABLE citations:
     • Explaining concepts (e.g., "establishing provenance chains [citations]")
     • Justifying methodology (e.g., "addressing concerns about hallucination [citations]")
     • Future possibilities (e.g., "for future integration with standards [citations]")
     • Background context or motivation
   • Rationale: Results reports OBSERVATIONS and MEASUREMENTS. Explanations belong in Introduction; justifications and future directions belong in Discussion.
   • Action if violated: Flag as minor issue. Remove explanatory citations or move content to Discussion if it describes future directions or broader implications.

Additional Critique Criteria

For Outlines (manuscript_plan.md):

1. Structure:

   • Does outline follow target journal format?
   • Are sections in logical order?
   • Is there clear progression from introduction to conclusion?

2. Evidence Mapping:

   • Is every claim linked to a specific data file?
   • Are figure references appropriate for each section?
   • Are code/script citations accurate?

3. Completeness:

   • Are all required sections present?
   • Is word count guidance realistic?
   • Are dependencies between sections clear?

For Literature Reviews:

1. Coverage:

   • Are foundational papers included?
   • Are recent advances (last 2 years) covered?
   • Are all major competing methods discussed?

2. Balance:

   • Is there appropriate coverage across different approaches?
   • Are strengths AND limitations discussed?
   • Is the positioning of the manuscript work clear?

3. Accuracy:

   • Can all citations be verified?
   • Are author names, years, venues correct?
   • Are paper summaries accurate (not hallucinated)?

4. Integration:

   • Are there clear suggestions for where to cite papers?
   • Does it identify gaps the manuscript addresses?
   • Is there guidance for updating existing sections?

Prose Linting (For Manuscript Drafts)

Run the prose linter: python scripts/rrwrite-lint-manuscript.py {target_dir}/full_manuscript.md

Output Format (per schema: schemas/manuscript.yaml)

Generate a critique report in {target_dir}/ directory with naming convention:

• Manuscript: {target_dir}/critique_manuscript_v1.md (increment version number for subsequent critiques)
• Outline: {target_dir}/critique_outline_v1.md
• Literature: {target_dir}/critique_literature_v1.md
• Section: {target_dir}/critique_section_v1.md

Filename pattern: critique_TYPE_vN.md where TYPE is (outline|literature|section|manuscript) and N is version number

Structure:

# Critique: [Document Type]

**Critiqued:** [Date]
**Document:** [File path]
**Target Journal:** [If applicable]

## Summary Assessment
[2-3 sentence overall evaluation]

## Strengths
1. [Positive aspect 1]
2. [Positive aspect 2]
3. [Positive aspect 3]

## Major Issues
1. **[Issue Category]:** [Description]
   - **Impact:** [Why this matters]
   - **Action:** [Specific fix required]

2. **[Issue Category]:** [Description]
   - **Impact:** [Why this matters]
   - **Action:** [Specific fix required]

## Minor Issues
1. **[Issue Category]:** [Description]
   - **Action:** [Quick fix]

2. **[Issue Category]:** [Description]
   - **Action:** [Quick fix]

## Compliance Checklist
[For manuscripts only]
- [ ] Abstract word count (if applicable)
- [ ] Citations verified
- [ ] Figures referenced
- [ ] Data availability statement
- [ ] Ethics statement (if needed)

## Actionable Next Steps
1. [Specific instruction for fixing issue 1]
2. [Specific instruction for fixing issue 2]
3. [Specific instruction for fixing issue 3]

## Recommendation
[ ] Accept with minor revisions
[ ] Major revisions required
[ ] Reject - fundamental issues

Validation

After generating the critique, validate it:

python scripts/rrwrite-validate-manuscript.py --file {target_dir}/critique_TYPE_vN.md --type critique

State Update

After successful validation, update workflow state with critique iteration:

import sys
from pathlib import Path
sys.path.insert(0, str(Path('scripts').resolve()))
from rrwrite_state_manager import StateManager

manager = StateManager(output_dir="{target_dir}")

# Get version number from filename or use manager method
critique_type = "manuscript"  # or "outline", "literature", "section"
version = manager.get_next_critique_version(critique_type)

# Count issues from the critique file
major_issues = 0  # Count from "MAJOR:" sections in critique
minor_issues = 0  # Count from "MINOR:" sections in critique
recommendation = "MAJOR_REVISIONS"  # Extract from recommendation section

manager.add_critique_iteration(
    critique_type=critique_type,
    version=version,
    file_path=f"{target_dir}/critique_{critique_type}_v{version}.md",
    recommendation=recommendation,
    major_issues=major_issues,
    minor_issues=minor_issues
)

Display updated progress:

python scripts/rrwrite-status.py --output-dir {target_dir}

Report validation status, critique iteration, and updated workflow progress. If validation passes, confirm critique completion.