aws-setup
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill extensively uses the AWS CLI to manage cloud infrastructure, including provisioning OpenSearch domains and Serverless collections, creating security policies, and managing IAM roles. It also executes a local Python script
scripts/opensearch_ops.pyto launch a management UI. - [EXTERNAL_DOWNLOADS]: Fetches and runs MCP servers using
uvxfrom trusted sources (such asawslabs), vendor-specific packages (opensearch-project), and well-known service domains (api.aws). These external components are necessary for the skill's intended functionality of managing AWS and OpenSearch resources. - [INDIRECT_PROMPT_INJECTION]: The skill configures agentic search capabilities that process external user data. It establishes an attack surface for indirect prompt injection by setting up conversational and flow agents. However, the instructions include proper boundary markers such as system prompts and specific template variables (
${parameters.user_prompt}) to mitigate these risks. The capabilities granted are scoped to OpenSearch and Bedrock operations required for search functionality.
Audit Metadata