federate-lakehouse-catalog

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill guides the agent to execute a series of sensitive administrative commands using the gcloud and aws CLI tools. These commands are used to enable APIs, create secrets in Secret Manager, provision BigLake catalogs, and modify IAM policies across both Google Cloud and AWS environments.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It instructs the agent to collect variable configuration data from the user (e.g., project IDs, catalog names, AWS role names) and directly interpolates this untrusted data into shell command strings. This lack of sanitization could allow a malicious user to perform command injection if the execution environment is not properly isolated.
  • Ingestion points: Configuration variables collected from the user in 'Step 1: Information Gathering'.
  • Boundary markers: Absent; there are no instructions to the agent to ignore or delimit potentially malicious payloads within the user-provided configuration values.
  • Capability inventory: Execution of gcloud and aws shell commands with administrative permissions.
  • Sanitization: Absent; the skill does not provide validation logic or escaping for the user-supplied strings before they are used in commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 11:51 AM
Security Audit — agent-trust-hub — federate-lakehouse-catalog