deploy-verification
Warn
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses command-line interfaces including kubectl, gh, and pulumi to interact with infrastructure and verify deployment status. These tools are used to query state and provide evidence of rollouts.
- [DATA_EXFILTRATION]: The skill instructs the agent to access sensitive information, such as container environment variables and mounted configuration files, using kubectl exec. This is explicitly intended to confirm secret and config values, which exposes internal process configuration to the agent context.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted output from external systems, creating a surface for indirect prompt injection where malicious instructions could be embedded in the data the skill ingests.
- Ingestion points: Output from gh, kubectl, and pulumi commands, as well as Prometheus metrics results in SKILL.md.
- Boundary markers: None defined to isolate or ignore embedded instructions in the retrieved output.
- Capability inventory: Shell command execution via kubectl, gh, and pulumi.
- Sanitization: No sanitization or validation of the external command output is specified.
Audit Metadata