academic-paper

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses standard academic tools for document processing. The Formatter Agent and Visualization Agent instruct the system to use 'pandoc' for Word/PDF conversion, 'xelatex' and 'tectonic' for LaTeX compilation, and 'python' or 'R' for generating statistical figures. These are standard tool invocations for the intended purpose of academic document preparation.
  • [EXTERNAL_DOWNLOADS]: The Citation Compliance Agent references the Retraction Watch Database (http://retractionwatch.com) and various official institutional domains (e.g., ndc.gov.tw, moe.edu.tw) for data verification. These are well-known, reputable services and organizations used for legitimate academic integrity checks.
  • [DYNAMIC_EXECUTION]: The Visualization Agent generates Python (matplotlib/seaborn) and R (ggplot2) code based on study results to produce publication-quality charts. This dynamic code generation follows predefined templates and is a core functional feature for academic reporting.
  • [DATA_EXPOSURE]: The pipeline includes a 'Style Calibration' feature that ingests user-provided past papers. This data enters the agent context to learn the author's writing voice. The skill uses structured 'Material Passports' and 'Knowledge Isolation Directives' to manage data boundaries and prevent accidental leakage of information into unrelated tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 02:03 PM
Security Audit — agent-trust-hub — academic-paper