gcp-to-aws
Warn
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: In the discovery phase (references/phases/discover/discover.md), the skill creates a Python or Node.js script locally to process billing data and then executes it. This behavior is used to avoid loading large billing files into the LLM context but involves runtime code generation and execution.
- [COMMAND_EXECUTION]: The skill performs several command-line operations. It uses bash to inspect files, runs terraform commands (fmt, init, validate) to check generated infrastructure code (references/shared/terraform-validation.md), and attempts to open an HTML report using open or xdg-open (references/phases/generate/generate-artifacts-report.md).
- [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it ingests and processes untrusted data from the user's project directory, including Terraform configurations, application source code, and billing exports. This data is used to determine architecture designs, cost estimates, and to generate deployment scripts. There are no explicit sanitization steps described for handling this external content within the agent's internal reasoning.
- [EXTERNAL_DOWNLOADS]: The skill references several external resources, including documentation on official AWS domains and a feedback survey at pulse.amazon. These are recognized as well-known or trusted services.
Audit Metadata