codex-claude-academic-skills
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant ingestion surface as it processes untrusted data from user-provided PDF research papers, MATLAB data files, and CSV datasets for classification and analysis. This creates a potential vulnerability where embedded malicious instructions or adversarial content within these files could influence the agent's logic during the summarization, writing, or code generation phases.
- [COMMAND_EXECUTION]: The skill involves the generation and execution of diverse command sets including shell commands for installation and configuration, Python scripts for document manipulation (Word/PPTX) and data analysis, and MATLAB/Octave scripts for signal processing simulations.
- [REMOTE_CODE_EXECUTION]: The skill provides instructions to download and install its components from an external GitHub repository and interacts with academic metadata services (such as doi.org and api.crossref.org) using the Python
requestslibrary to fetch and process remote citation data.
Audit Metadata