snapshot
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill contains numerous examples of shell commands for low-level system administration, including volume management (lvcreate, lvconvert, lvremove), filesystem operations (mount, btrfs, zfs), and service management (systemctl stop application). These operations require high privileges and are capable of significantly altering system state.
- [INDIRECT_PROMPT_INJECTION]: The bash script templates in the 'Common Scenarios' section (Scenario 1 and 2) use variables and shell piping (xargs) to process data like snapshot names. If an agent populates these templates using untrusted input without proper sanitization, it could be vulnerable to command injection.
- Ingestion points: User-provided volume IDs, disk IDs, and snapshot names used in bash examples and cloud CLI commands.
- Boundary markers: None present in the instruction set or script templates.
- Capability inventory: Execution of system administration tools (LVM, Btrfs, ZFS) and cloud provider CLIs (AWS EC2, Aliyun ECS).
- Sanitization: No input validation or escaping logic is provided in the example scripts.
Audit Metadata