mcpField
S
sci-review

sci-review

github.com/shzhao27208Documents213241 installsUpdated 11w ago
Repository

About sci-review

A writing skill for producing structured literature reviews and professional responses to peer reviewers. It provides recommended section frameworks, evidence-focused practices, and a lightweight validator for required headings and banned phrases.

Key features

  1. 01Four-part literature review structure
  2. 02Reviewer-response framework
  3. 03Evidence-focused writing guidance
  4. 04Lightweight output validator
  5. 05Golden-case validation tests

Use cases

  1. 01Drafting a structured literature review
  2. 02Preparing point-by-point reviewer responses
  3. 03Checking required sections and banned phrases

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/shzhao27208/aut_sci_write --skill sci-review

After installing, type/sci-reviewin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

_meta.json
references/
rebuttal-guide.md
section-guide.md
word-choice.md
scripts/
validate_review_output.py
templates/
literature-review-template.md
rebuttal-template.md
tests/
golden_cases.json

Details

From the project README

Sci-Review

Use this skill to produce structured literature-review writing and professional reviewer responses.

Literature Review Structure

Use this four-part structure unless the user requests a different journal format:

  1. Introduction: background, problem definition, gap identification, and contribution.
  2. Methodology: taxonomy, method classes, comparison dimensions, and performance evidence.
  3. Challenges: phenomenon, cause, and direction. Make the problem visible before proposing a route forward.
  4. Conclusion: distilled insights and a future roadmap.

Prefer specific evidence over broad claims. Replace vague phrases such as "significantly better" with measured comparisons when data is available.

Rebuttal Structure

For each reviewer point, use:

  1. Reviewer concern: restate the concern accurately and neutrally.
  2. Response: answer with evidence, clarification, or a limitation acknowledgement.
  3. Revision plan: state the exact manuscript change, including section, table, figure, appendix, or experiment when possible.

Avoid adversarial phrasing such as "reviewer misunderstood" or "the reviewer is wrong". Use constructive language such as "we will clarify this point in the manuscript" or "we agree that additional evidence would improve the presentation".

Validation

The skill includes a lightweight validator. Run from the skills/sci-review/ directory:

# From the skills/sci-review/ directory:
python scripts/validate_review_output.py --case literature-review --output output.md
python scripts/validate_review_output.py --case rebuttal --output output.md
python scripts/validate_review_output.py --list-golden

The validator checks required section names and banned phrases. Golden cases live in tests/golden_cases.json; they define expected output features rather than exact wording.

Best Practices

  • Read the source literature, reviewer comments, or draft before rewriting.
  • Preserve technical nuance. Do not invent experiments, results, baselines, or citations.
  • Mark uncertainty explicitly when source evidence is missing.
  • Keep tone professional, direct, and evidence-driven.

Comments

Markdown supported