1anj/academic-integrity-skill

This is an academic integrity self-assessment skill designed for scenarios such as submission, revision, post-publication questioning, and internal review of research groups, including Nature. This is an executable workflow used to screen papers for data falsification, image manipulation, statistical anomalies, methodological gaps, citation and author declaration issues, and to generate self-inspection reports and corrective plans.

This SKILL is positioned as “self-inspection and self-correction” and “evidence-based risk grading,” and does not replace academic misconduct determinations made by journals, institutions, or ethics committees.

File Structure

Applicable Scenarios

• Conduct internal academic integrity self-checks before submitting to journals including Nature/Nature Portfolio.
• Conduct evidence-based reviews after receiving concerns from editors, reviewers, PubPeers, or readers.
• Check for possible data falsification, image reuse, Western blot splicing, statistical anomalies, p-hacking, duplicate publication, or citation issues.
• The PI, corresponding authors, and co-authors of the research group organize the facts before submitting revisions, retraction notes, or institutional investigation materials.
• Conduct preliminary statistical screening of the CSV/XLSX values extracted from the chart.

Review Rules

1. Scientific research images and image completeness

This skill is not just about numerical checks. Image review is a core process, focusing on checking:

• Image reuse: Whether the same image or a local area is used in different samples, different treatment groups, different time points, different magnifications, different stains/channels, different papers, or supplementary images.
• Reuse after transformation: Rotate, flip, zoom, stretch, crop, adjust contrast, change colors, or partially mask to see whether the same cells, tissue structure, background texture, dust, bubbles, scratches, or noise patterns are still visible.
• Western blot / gel problem: Repeated use of internal reference strips; Different experiments share the same beta-actin/GAPDH/tubulin; Swimlane shape and height are the same; Non-adjacent lanes are not spliced together; There are vertical boundary lines between lanes; Background grayscale abrupt change; Inconsistent exposure; Missing molecular weight markers; Cropped images cannot be traced back to the uncropped original image.
• Microscopic Pattern Issues: Different groups show the same field of view, same cell population, and same tissue structure; Local cloned cells or tissue regions; Inconsistent channel superposition; Selective thresholds, denoise, gamma, pseudo-color, deconvolution, or representative field selection are not explained.
• Flow Graph Problem: Scatter cloud duplication, gated line duplication, quadrant scale not matching point cloud in the graph, missing compensation/gating hierarchy, and the same isotype/FMO/control used for irrelevant experiments.
• Migration, Invasion, Scratching, Clone Formation, Animals and Histology Maps: Duplicate cloning, repeated scratch edges, repeated tissue textures, duplicated animal images, inconsistent scales, mismatched field of view with legend.
• Chart manipulation: Bar charts/line charts do not match source data; Y-axis does not start from zero and exaggerates effects; Scale spacing is unequal; Unit is missing; Error bar definitions are missing; Error bars are too small to meet experimental noise; Data points or error bar shapes in different charts are duplicated.
• Composite Chart Disclosure: All cropping, stitching, non-adjacent lane parallels, overall brightness/contrast adjustments, representative image selection, false color, and threshold handling should be explained in the legend or method.