startup-it-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes high-privilege shell commands (using
sudo) to perform system-level tasks including flushing DNS caches, managing network interfaces, and restarting system services (e.g., CUPS, print spoolers). - [PRIVILEGE_ESCALATION]: Instructions are provided for modifying sensitive system security settings, such as enabling Remote Desktop on Windows, configuring firewall states on macOS and Linux, and managing disk encryption via BitLocker and FileVault.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various external administrative interfaces, creating an injection surface.
- Ingestion points: Reads outputs from Google Workspace (via GAM), Okta Lifecycle APIs, Slack Admin APIs, GitHub Member APIs, and AWS IAM commands (SKILL.md).
- Boundary markers: Absent. There are no delimiters or warnings provided to prevent the agent from interpreting instructions potentially embedded in the data returned by these tools.
- Capability inventory: High capability across all scripts, including user suspension, credential revocation, session termination, network interface modification, and local file writes.
- Sanitization: None. Data from external APIs is used directly in workflows without validation or escaping.
Audit Metadata