automated-review-guide

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious behavior or indicators of compromise were found. The skill focuses on legitimate academic review workflows.\n- [PROMPT_INJECTION]: The provided prompts are task-specific instructions for reviewing manuscript sections (Abstract, Methods, Results). They do not contain instructions to override model constraints or bypass safety protocols.\n- [DATA_EXFILTRATION]: No network operations or unauthorized data access patterns are present. The documentation explicitly advises against uploading confidential manuscripts to untrusted services, demonstrating a security-conscious approach.\n- [COMMAND_EXECUTION]: The included Python script check_statistical_reporting is limited to text processing using the standard re module. It performs no external command execution, file system manipulation, or network communication.\n- [PROMPT_INJECTION]: (Indirect Surface) The skill involves processing external manuscript data. \n
  • Ingestion points: Prompts are designed to analyze user-provided text from papers (SKILL.md).\n
  • Boundary markers: Prompts use clear descriptive labels but lack explicit delimiters for raw input.\n
  • Capability inventory: The skill has no dangerous capabilities such as file writes or shell execution.\n
  • Sanitization: No sanitization is performed, which is acceptable for the intended use-case as an advisory guide.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 10:16 PM