domains / research / scientific-manuscript-writing / scientific-writing

scientific-writing

Skill card

combined

The copies overlap on IMRAD structure, citation styles, and reporting guidelines, but they disagree on scope and mandatory steps. K-Dense-AI/scientific-agent-skills focuses on evidence provenance, confidentiality, and local consistency checks without requiring visual generation. foryourhealth111-pixel/Vibe-Skills explicitly excludes generating visual assets. brycewang-stanford/Auto-Empirical-Research-Skills mandates the generation of graphical abstracts and multiple figures using specific scripts. sickn33/agentic-awesome-skills is a thin wrapper that defers to a detailed guide. Because one copy forbids what another mandates, and the others provide unique procedural constraints (evidence binding vs. visual generation), no single copy is the authentic procedure for all.

Repositories

Other copies

Combined skill

# Scientific Writing

## Purpose
Produce clear scientific prose using IMRAD structure, citation-integrated paragraphs, and reporting guidelines. Maintain scientific accuracy and traceability.

## Workflow
1. **Intake & Structure**: Identify document type, study design, and target venue. Select the appropriate reporting guideline (e.g., CONSORT, PRISMA).
2. **Evidence & Provenance**: 
   - Map factual claims to verified evidence IDs.
   - Maintain a source manifest and claim registry to ensure traceability.
   - Do not fabricate citations, data, or results.
3. **Drafting**: 
   - Use a two-stage process: create section outlines, then convert to flowing prose.
   - Write in full paragraphs; avoid bullet points in the final manuscript.
   - Apply the selected citation style (APA, AMA, Vancouver, etc.).
4. **Visuals & Figures**:
   - Write clear captions and in-text callouts for figures and tables.
   - *Conflict Note*: Depending on the specific tooling available, you may be required to generate graphical abstracts and schematics (as per brycewang-stanford) or restricted to writing captions only (as per foryourhealth111-pixel). Check local tool availability before attempting generation.
5. **Review & Submission**: 
   - Check for consistency in units, denominators, and sample sizes.
   - Ensure all declarations (funding, conflicts, AI use) are present and accurate.
   - Verify confidentiality and authorization for any external processing.

## Safety & Ethics
- **No Fabrication**: Never invent citations, data, or authorship details.
- **Confidentiality**: Do not send unpublished or sensitive data to external services without explicit authorization and policy review.
- **Scientific Fidelity**: Preserve uncertainty, distinguish association from causation, and report negative or null findings when appropriate.