cobol-migration

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. Ingestion points: Raw COBOL source code is ingested via the parse_cobol method in __init__.py. Boundary markers: The SKILL.md does not define specific delimiters or instructions to ignore potential commands embedded in COBOL comments. Capability inventory: The agent is empowered to perform code generation (CodeTransformation) and test synthesis (TestSynthesis) based on the input. Sanitization: There is no evidence of logic to sanitize or strip comments from the legacy code before LLM processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 06:34 AM