infrastructure-as-code
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill generates infrastructure code based on untrusted user requirements, which could be exploited to provision insecure resources if the execution plan is not carefully reviewed.
- Ingestion points: User-provided cloud specifications and resource requirements described in the workflow section of SKILL.md.
- Boundary markers: The workflow explicitly includes a manual review step ("Execute Plan and Apply") where the agent must present the plan for user approval before provisioning.
- Capability inventory: The skill uses tools like Terraform and Pulumi to perform file writes and execute command-line operations (plan, apply, preview) that interact with cloud provider APIs.
- Sanitization: There is no explicit input sanitization mentioned; the skill relies on the human reviewer to identify malicious or incorrect infrastructure plans.
- [COMMAND_EXECUTION]: The skill facilitates the execution of powerful infrastructure management commands (terraform apply, pulumi up) that can create, modify, or delete cloud resources.
- [PRIVILEGE_ESCALATION]: The skill provides examples for creating IAM roles and attaching policies (e.g., AmazonDynamoDBFullAccess in Example 2), which involves managing permissions within the cloud environment.
- [EXTERNAL_DOWNLOADS]: The skill references and uses external infrastructure modules, such as the terraform-aws-modules/vpc/aws module from the Terraform Registry.
- The module is a well-known community standard and is pinned to a specific version (5.1.0).
Audit Metadata