food-label-review

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill files were scanned for malicious patterns including prompt injection, data exfiltration, and obfuscation. All references are static informational Markdown files used for compliance checking.\n- [COMMAND_EXECUTION]: The skill utilizes a Node.js script scripts/generate_report.js to process audit results into a Word document. This script is a legitimate vendor resource that performs local file system operations without external network interaction.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-submitted food labels, which represents an ingestion surface. \n
  • Ingestion points: User-provided food label text or images processed during the audit workflow in SKILL.md.\n
  • Boundary markers: None explicitly defined in the prompt instructions to separate user content from system logic.\n
  • Capability inventory: The agent can execute a local reporting script to create file output.\n
  • Sanitization: No explicit sanitization of user-provided label text is performed before it is included in the generated report.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 03:46 AM