Paper CT Seven-Dimensional Flaw Scan System

Problem

Before submission, papers often fail on formatting rules, weak evidence or citations, and AI-generated stylistic traces. Manual review is slow and easy to miss, especially when Chinese academic papers must follow scattered requirements such as GB/T 7714 references, figure/table numbering, funding disclosure, and journal fit.

How It Works

The skill follows a CT workflow: penetrate → diagnose → prescribe. After a trigger such as “check this paper,” it reads the local paper file and produces:

• output/report_summary.md: a human-readable report
• output/report_detail.json: structured findings for downstream processing
• output/scores.json: scores, total, and grade

Coverage spans seven weighted dimensions: content authenticity 25%, feasibility 15%, truthfulness 15%, AI traces 15%, layout compliance 10%, academic norms 10%, and journal fit 10%. It does more than list issues; it points to repair actions. Failures are not silently skipped—if a module stops, the report marks that dimension while other checks continue.

Boundaries

It targets Chinese academic papers and runs locally without uploading content. It does not perform plagiarism checking, English polishing, authorial rewriting, or guarantee acceptance, and it is not intended for non-academic texts such as contracts or fiction. Pair it with dedicated grammar tools or commercial plagiarism systems when needed.