sealos-deploy
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates within a managed sandbox environment using platform-injected credentials (KUBECONFIG, GITHUB_TOKEN) for legitimate deployment tasks. It explicitly forbids logging or printing sensitive environment variables or user inputs.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data, creating an attack surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the agent context through SEALAI_INPUTS_PATH (user-form responses) and the contents of the workspace repository being deployed.
- Boundary markers: The instructions include explicit security constraints forbidding the logging of user inputs or secrets and require using the --args-file flag to pass inputs to the API script to prevent interpolation into prompt strings.
- Capability inventory: The skill has the capability to interact with the Kubernetes API via kubectl and execute local Python scripts (kaniko-build.py, sealos-api.py) for builds and deployments.
- Sanitization: The workflow relies on the sibling sealos-api.py script to safely handle and parse user-supplied arguments from the designated inputs file.
Audit Metadata