labarba/sciwrite

What it solves

SciWrite is designed to help graduate students and researchers improve the clarity and quality of scientific and engineering manuscripts. It addresses the common problem where authors often struggle to self-edit for clarity because they are too close to the same text, and often lack formal training in scientific writing.

How it works

The project provides an "Agent Skill" (a SKILL.md file) that encodes a systematic editorial review process based on Dr. Kristin Sainani's Writing in the Sciences course. When loaded into an AI tool (like Claude, ChatGPT, or Gemini), the AI follows five sequential audit passes:

1. Clutter extraction: Removes dead-weight phrases and filler language.
2. Voice and verb vitality: Replaces passive constructions and "smothered verbs" (nominalizations) with active alternatives.
3. Sentence architecture: Evaluates length, structure, and logical flow.
4. Keyword consistency: Ensures technical terms are used consistently to avoid ambiguity.
5. Numerical and citation integrity: Checks for internal consistency of values and citations.

It supports four review modes: full-manuscript, single-section, targeted (specific pass), and interactive (paragraph-by-paragraph).

Who it’s for

Researchers, graduate students, and engineers who are writing scientific manuscripts and want to professional editorial feedback based on a specific, proven methodology.

Highlights

• Methodology-driven: Based on a Stanford University course on scientific writing.
• Content-preserving: Focuses on improving delivery without altering scientific data or technical claims.
• Crosspatfrom compatible: Works across major AI assistants via the SKILL.md standard (Claude, ChatGPT, Gemini).
• Structured output: Provides a report with specific findings, concrete revisions, and severity tags.