aws-iam
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Analysis of untrusted inputs: The skill analyzes user-provided source code and Terraform plans. This constitutes an indirect prompt injection surface. The skill mitigates this by using a deterministic tool from a trusted vendor rather than relying solely on LLM interpretation.
- External tool execution: The skill uses
uvxto execute an analysis tool. This involves downloading and running code from a remote source. The source is a trusted organization and the tool is standard for IAM policy automation. - Metadata discovery from sensitive files: The skill identifies AWS account IDs and regions by reading files like
.envandterraform.tfvars. While these files often contain secrets, the skill specifies that only non-sensitive metadata should be extracted to improve the specificity of generated policies. - Remote data retrieval: The skill fetches IAM service authorization details from official AWS endpoints. This ensures the agent uses current and verified mapping data when constructing policies.
Audit Metadata