upgrading-cluster-version
Warn
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill requires the use of elevated privileges via
sudoto stop and start database services and modify system-level binaries (e.g.,sudo systemctl stop cockroachdbandcp /path/to/new/cockroach /usr/local/bin/cockroach). This allows for modification of the host operating system's service states. - [COMMAND_EXECUTION]: The instructions involve significant command-line interaction, including binary management, cluster state validation through the
cockroachCLI, and Kubernetes resource orchestration usingkubectlandhelm. - [INDIRECT_PROMPT_INJECTION]: The skill routes its logic based on user-provided deployment context, such as the tier and version strings. These inputs represent an attack surface where untrusted data influences the instructions provided by the agent.
- Ingestion points: User-provided deployment tier and current version strings specified in the context gathering section of
SKILL.md. - Boundary markers: None identified in the skill instructions to separate user input from the rest of the prompt context.
- Capability inventory: Includes system service management, file system writes in privileged paths, Kubernetes API modifications, and network requests to cloud management APIs.
- Sanitization: No validation or sanitization of the version or tier input is described.
Audit Metadata