agent-platform-endpoint-management
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill leverages the
gcloudCLI to manage cloud infrastructure. It provides specific command templates for interacting with Google Cloud's AI Platform services, including creating, updating, and deleting endpoint resources. - [INDIRECT_PROMPT_INJECTION]: The skill uses variables such as
$PROJECT_ID,$LOCATION_ID, and$ENDPOINT_IDwithin shell commands, creating a potential surface for injection if user-supplied values are not handled correctly. However, this is mitigated by strict instruction tiers (Tier M and Tier D) that mandate the agent must wait for explicit, typed user confirmation and display the exact literal command before any execution occurs.
Audit Metadata