aws-lambda-microvms
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Isolated Code Execution Surface: The skill describes building and running applications within Firecracker-isolated MicroVMs. This is the primary function of the service, designed for workloads requiring strong tenant separation, such as AI agent sandboxes, interactive notebooks, and security scanners. The isolation provided by Firecracker is a key security feature for these use cases.
- Interactive Shell Capability: The documentation details the use of the
SHELL_INGRESSnetwork connector to enable programmatic shell access for debugging or agent workflows. The skill includes guidance on managing shell auth tokens, explicitly advising users to treat them as secrets and avoid logging or storing them in insecure locations. - Indirect Prompt Injection Surface: The skill acknowledges the use of MicroVMs for executing user-supplied code in AI agent environments, which presents a surface for indirect prompt injection. The documentation frames Firecracker microVM isolation as the primary mitigation for this risk, providing a fresh and isolated environment per session.
- Dependency and Resource Management: The provided examples use standard, versioned Python packages (
flask==3.0.3) and official AWS base images from public ECR. The skill explains the separation of build and execution roles to maintain least-privilege access to S3 artifacts and runtime resources.
Audit Metadata