analyzing-schema-change-storage-risk

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the use of user-supplied node addresses and table names. Ingestion points: user-provided values for the <node> placeholder in curl commands and <table> in SQL queries in SKILL.md. Boundary markers: The instructions lack boundary markers or warnings to the agent to disregard embedded instructions. Capability inventory: The skill utilizes shell command execution via curl and database metadata access via SQL. Sanitization: There is no evidence of input validation or sanitization for these parameters.
  • [COMMAND_EXECUTION]: The skill directs the agent to execute curl commands against user-defined node addresses.
  • [DATA_EXFILTRATION]: The skill employs the insecure -k flag with curl, which may expose internal metrics to interception. Additionally, as noted in the provided documentation, SHOW RANGES WITH DETAILS can expose primary key data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 03:58 PM
Security Audit — agent-trust-hub — analyzing-schema-change-storage-risk