NYC

relational-database-mcp-cloudbase

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • Indirect Prompt Injection (HIGH): The skill creates an attack surface where data read from the database via executeReadOnlySQL can influence subsequent write or DDL operations.
  • Ingestion points: Results from executeReadOnlySQL and readSecurityRule in SKILL.md.
  • Boundary markers: Absent; there are no instructions for the agent to use delimiters or ignore instructions found within retrieved data.
  • Capability inventory: The agent has high-privilege access via executeWriteSQL (INSERT, UPDATE, DELETE, DROP) and writeSecurityRule (modifying database permissions).
  • Sanitization: Absent; the skill does not mandate sanitization or parameterization of data before it is used in SQL statements.
  • Dynamic Execution (MEDIUM): The skill's primary function is the execution of arbitrary SQL commands through executeWriteSQL, which is a form of dynamic code execution that can be exploited if the inputs are not strictly controlled.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 09:09 PM