system-field-reference

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill directs the agent to utilize a command-line tool, ae-cli, to query metadata and create virtual properties within the analytics engine environment. This allows the agent to perform administrative tasks based on the provided instructions.
  • [INDIRECT_PROMPT_INJECTION]: User-provided field names and display names are interpolated directly into shell-based CLI commands. An attacker could potentially provide a malicious field name to attempt to escape the command arguments or manipulate the resulting SQL expression.
  • Ingestion points: User input captured during the 'Clarify User Requirements' phase, specifically requested field names and naming preferences used in Step 4 of the workflow.
  • Boundary markers: The instructions advise the agent to quote system-field identifiers containing special characters like #, $, or @ within the SQL expression argument.
  • Capability inventory: The skill uses ae-cli to perform listing and creation operations on the project's metadata gateway, which influences how data is displayed and analyzed in the frontend.
  • Sanitization: The skill relies on natural language instructions for the agent to apply quoting to identifiers; it does not include programmatic sanitization or strict validation logic for the user-supplied strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 03:22 AM
Security Audit — agent-trust-hub — system-field-reference