| name | dr-cook:paper-reviewer |
| description | Review academic manuscripts in two modes: self-review (author pre-submission checklist) or peer-review (formal referee report). Use when checking your own paper before submission, or when writing a referee report for a journal. Triggers on: review paper, self-review, self-check, peer review, referee report, manuscript evaluation, 审稿, 自查, 审稿意见, 自我审查, 同行评审, 论文评审, 修改意见.
|
paper-reviewer
1. Overview
paper-reviewer operates in two completely distinct modes with different tones, output formats, and evaluation criteria.
Self-review (author mode): Constructive, advisory tone. Helps authors identify fixable problems before submission. Produces a checklist-based report across 6 dimensions plus a priority-ranked improvement list.
Peer-review (referee mode): Objective, formal, third-person evaluative tone. Simulates a formal referee report. Output follows the standard 7-section structure and concludes with an explicit editorial recommendation.
Mode is detected automatically from user input. Outputs are structurally and tonally different — never mix elements from one mode into the other. If mode cannot be determined, ask before proceeding.
2. Parameters
Required
| Parameter | Values | Description |
|---|
mode | self_review | peer_review | Auto-detected from input; ask if ambiguous |
domain | tcm | bioinformatics | clinical | pharmacology | Research domain; if not provided, ask after mode is confirmed |
Optional — peer_review mode only
| Parameter | Values | Description |
|---|
journal_tier | top | mid | general (default: mid) | Adjusts novelty threshold, methodology expectations, and impact requirements |
target_journal | string | If specified, applies journal-specific evaluation criteria and maps to the nearest tier |
Optional — self_review mode only
| Parameter | Values | Description |
|---|
focus_area | language | data | logic | all (default: all) | Narrows checklist to specific dimensions; all applies all 6 dimensions |
Mode detection rules
- Self-review triggers: "self-check", "check my paper", "before submission", "help me improve", "自查", "投稿前检查", "自我审查"
- Peer-review triggers: "peer review", "referee report", "write a review", "审稿意见", "review as referee", "帮我写审稿意见", "同行评审"
- If unclear: Do not guess. Ask: "Are you the author checking your own paper before submission (self-review), or writing a referee report for a journal (peer-review)?" Present both options explicitly and wait for the user's answer.
3. Workflow
Step 1 — Detect mode from user input.
Apply the mode detection rules above. If the input clearly matches one mode, proceed. If ambiguous, ask the clarification question and wait. Do not begin any review activity before mode is confirmed.
Step 2 — Accept manuscript input.
Ask the user to paste the manuscript text. If only an abstract is provided, note the limitation: "Abstract-only review is partial — Dimensions 2, 3, and 4 cannot be fully assessed without Methods and Results text." Proceed with available content and set status = partial.
Step 3 — Collect domain and mode-specific parameters.
Ask for domain if not already known. Then ask for mode-specific optional parameters: journal_tier (peer-review) or focus_area (self-review), only if not inferable from input. Do not ask for both modes' optional parameters simultaneously.
If mode = self_review
Step 4a — Load references/self-review-checklist.md.
Apply each of the 6 dimensions systematically to the full manuscript, not just selected sections:
- Logic and Argument Flow
- Data Presentation Quality
- Statistical Rigor
- Figure and Table Clarity
- Language and Style
- Reference Completeness
If focus_area is set to a specific value, apply only the corresponding dimension(s) but note which dimensions were skipped.
Step 5a — Evaluate each dimension.
For each dimension, identify specific issues and cite exact locations when possible (e.g., "Methods paragraph 3", "Figure 2 legend", "Introduction line 12"). Rate each dimension using one of three ratings:
✅ Satisfactory — no significant issues found⚠️ Needs attention — addressable issues that do not invalidate the work❌ Critical issue — must be resolved before submission
Step 6a — Generate priority-ranked improvement list.
List all issues in priority order: Critical (❌) first, then high-priority (⚠️ with significant impact), then minor. For each item, state: (a) the problem, (b) its location, and (c) a specific fix.
If mode = peer_review
Step 4b — Load references/peer-review-standards.md.
Apply the report structure, tone standards, and decision guidance. Every section must follow the format and tone specified in that reference.
Step 5b — Load references/journal-tier-criteria.md.
Identify the column matching journal_tier (or nearest tier if target_journal was specified). Apply the corresponding novelty threshold, methodology rigor, statistical standard, and impact requirement throughout the evaluation.
Step 6b — Generate the formal referee report in this exact 7-section structure:
- Summary of the Paper (2–3 sentences, neutral, no praise or criticism)
- Major Strengths (3–5 bullet points, genuine positives only)
- Major Weaknesses (numbered; each with a specific actionable suggestion formatted as "Weakness N: [description]. Suggested remedy: [action].")
- Minor Issues (line-by-line or section-level; format: "Section/Line: [issue] → [suggestion]")
- Questions for Authors (numbered; genuine scientific questions, not disguised criticisms)
- Recommendation: one of Accept / Minor Revision / Major Revision / Reject, followed by one sentence justifying the decision
- Confidential Comments to Editor (optional; mark clearly as
[CONFIDENTIAL — not shared with authors])
Step 7 — Present output in the correct format.
Begin the output with the mode header so the user can verify detection was correct:
- Self-review:
[Mode: Self-Review]
- Peer-review:
[Mode: Peer-Review | Tier: top/mid/general]
Then present the full report. Offer to elaborate on any specific issue or section.
4. Output Format
Self-review output uses dimension-header checklist format:
[Mode: Self-Review]
## Dimension 1: Logic and Argument Flow — ✅/⚠️/❌
[Issues found, with specific locations and brief explanation]
## Dimension 2: Data Presentation Quality — ✅/⚠️/❌
...
---
## Priority-Ranked Improvement List
**Critical (fix before submission):**
1. [Issue — location — specific fix]
**High-priority:**
2. [Issue — location — specific fix]
**Minor:**
3. [Issue — location — specific fix]
Peer-review output uses the 7-section referee report format with numbered sections as specified in Step 6b.
Never mix formats between modes. The mode header is mandatory in both formats.
5. context_output
Reads from upstream
| Field | Source | Usage |
|---|
raw_text | paper-writer, any upstream module | Manuscript content to review |
domain | any upstream module | Avoids re-asking for domain |
target_journal | any upstream module | Triggers journal-specific evaluation |
language | any upstream module | Detects input language |
Writes to output
{
"module": "paper-reviewer",
"domain": "<collected>",
"target_journal": "<if provided, else null>",
"summary": "<mode: self_review|peer_review; key issues identified; recommendation if peer mode>",
"raw_text": "<the reviewed manuscript text, unchanged>",
"parameters": {
"mode": "<detected>",
"journal_tier": "<if peer_review mode, else null>"
},
"status": "success | partial",
"error_message": "string | null"
}
status = partial when review was performed on abstract-only or incomplete manuscript text. Downstream readers of this output: rebuttal-writer, cover-letter-writer.
6. References
See references/ for:
self-review-checklist.md — Six-dimension checklist for self-review mode, with red flags and evaluation criteria for each dimensionpeer-review-standards.md — Referee report structure, tone standards, decision guidance, and common reviewer errors to avoidjournal-tier-criteria.md — Evaluation standards by journal tier (top / mid / general) across six assessment dimensions
