sui-bcs

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): The skill demonstrates processing of untrusted external data which is subsequently used to drive high-privilege operations, constituting a significant Indirect Prompt Injection surface.
  • Ingestion points: Untrusted data enters the agent context via functions like RegistrationRequest.parse(requestData) and TransferRequest.parse(requestBytes) in reference/workflows.md.
  • Boundary markers: The skill relies on structural BCS schemas for data delimitation but lacks explicit instructions or delimiters to prevent embedded natural language commands within parsed strings from influencing downstream reasoning.
  • Capability inventory: The documentation examples include database write operations (saveUserToDatabase) and the construction and execution of blockchain transactions (tx.moveCall, tx.setGasBudget) in reference/workflows.md.
  • Sanitization: The skill provides robust examples of sanitization including cryptographic signature verification (validateSignature), timestamp checks to prevent replay attacks, and type-specific range validation for numeric and string types.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 05:54 AM