Summary

SciWrite is an AI agent skill designed to enhance the clarity and quality of scientific and engineering manuscripts. It addresses the common challenge of self-editing by systematically applying Dr. Kristin Sainani's established "Writing in the Sciences" methodology. Targeted at researchers and students, this skill provides an automated editorial review process, improving argument accessibility and overall readability without altering core scientific content.

How It Works

The skill encodes a rigorous editorial process derived from 30 lectures on scientific writing. It performs five sequential audit passes: identifying and replacing clutter, strengthening voice and verb usage (addressing passive voice and nominalizations), evaluating sentence architecture and flow, ensuring keyword consistency, and verifying numerical and citation integrity. Users can select from full-manuscript, section-specific, targeted single-pass, or interactive paragraph-by-paragraph review modes. This approach offers a systematic, consistent, and machine-actionable application of established writing principles.

Quick Start & Requirements

Integration involves using the SKILL.md file with compatible AI tools rather than a direct installation.

• Primary Usage: Load SKILL.md into AI assistants supporting the Agent Skill format.
• Supported Platforms: Claude.ai, Claude Code, ChatGPT (via Custom GPT), Google Gemini (via Gems), GitHub Copilot, Cursor.
• Prerequisites: Access to one of the supported AI tools. No specific coding or library dependencies are required for the skill itself.
• Documentation: Instructions and example prompts are available in the companion HOW-TO-USE.md file (within the repository).

Highlighted Details

• Cross-Platform Agent Skill: Implements the open Agent Skills standard (SKILL.md), ensuring compatibility across major AI platforms like Claude, ChatGPT, and Gemini.
• Established Methodology: Leverages Dr. Kristin Sainani's renowned Stanford "Writing in the Sciences" course, providing a robust, proven framework for technical writing improvement.
• Multi-Stage Audits: Conducts five distinct review passes (Clutter, Voice, Structure, Keywords, Numerics/Citations) for comprehensive writing quality assessment.
• AI-Assisted Workflow: Demonstrates a meta-workflow where AI tools (browser, synthesis engine, skill authoring) collaborate to create a new AI tool (the skill).

Maintenance & Community

Developed as part of a university course (MAE 6291, The George Washington University) and shared via the GW Engineering AI Academy. Contributions addressing edge cases or improving performance are welcomed via GitHub issues or pull requests.

Licensing & Compatibility

• License: CC BY 4.0 (Creative Commons Attribution).
• Permissions: Free to use, adapt, and share, requiring attribution.
• Commercial Use: Permitted, compatible with closed-source projects provided attribution is maintained. The underlying lecture material is also CC-BY licensed.

Limitations & Caveats

The skill may occasionally misinterpret highly specialized terminology or field-specific conventions, requiring user domain expertise for critical review of suggested revisions. Its effectiveness is dependent on the AI tool's implementation of the Agent Skill standard.