krt-deploy-summoner
Pass
Audited by Gen Agent Trust Hub on May 26, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill manages the use of standard infrastructure tools like
kubectl,helm, anddocker. It implements a safety-first model where diagnostic commands are preferred, and any action that modifies the environment (mutations) requires a clear plan and user authorization.\n- [EXTERNAL_DOWNLOADS]: The skill references official technical documentation from well-known and trusted sources including the Kubernetes project, Helm, and Docker for guidance on troubleshooting and best practices. No executable code or untrusted packages are downloaded.\n- [DATA_EXFILTRATION]: Instructions explicitly forbid the display or decoding of sensitive information such as secrets, environment variables, and authentication tokens, ensuring that diagnostic logs and reports remain secure.\n- [PROMPT_INJECTION]: The skill includes instructions to identify safety boundaries and requires explicit approval for remote mutations, which helps protect against instructions attempting to bypass security constraints. Regarding indirect injection:\n - Ingestion points: Container logs, Kubernetes events, and resource manifests (SKILL.md Step 3).\n
- Boundary markers: Structured planning format to separate evidence from actions.\n
- Capability inventory: Tool access includes
kubectl,helm, anddockerfor cluster management.\n - Sanitization: Relies on user verification of proposed command plans.
Audit Metadata