iam-helper-for-privileged-access-management
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill relies on the
gcloudCLI and local bash scripts to manage PAM entitlements and grants. These scripts perform resource hierarchy traversals to ensure that access is correctly scoped. This is the intended functionality for a cloud administration skill. - Indirect Prompt Injection: The skill processes output from Google Cloud commands (e.g., listing entitlements or describing resources). While this data is sourced from a trusted cloud provider, the skill includes a safety confirmation strategy where the agent must present a plan and obtain explicit user consent before executing any destructive or modifying actions.
- Safety Strategy Implementation: The skill explicitly instructs the agent to follow a 'Plan-Validate-Execute' pattern. This involves constructing proposed changes, validating them against corporate rules, and requiring a 'Yes/No' confirmation from the user for any creation, update, deletion, or approval task.
Audit Metadata