fujitsu-mainframe

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • 4. Unverifiable Dependencies & Remote Code Execution (SAFE): The skill utilizes the standard 'find' utility to search for source files locally. No remote scripts are downloaded or executed.
  • 8. Indirect Prompt Injection (SAFE): The skill analyzes external legacy code which could contain malicious instructions.
  • Ingestion points: Local filesystem access for COBOL, JCL, and configuration files.
  • Boundary markers: Absent in the current prompt instructions.
  • Capability inventory: File discovery and code generation.
  • Sanitization: No content filtering of analyzed source files is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:18 PM