pulumi-python
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to translate user requirements into cloud infrastructure-as-code (IaC) definitions. This interaction model creates a surface where malicious or poorly formed natural language instructions could lead to the creation of insecure cloud resources.
- Ingestion points: The agent receives infrastructure specifications and configuration parameters from users via natural language prompts (e.g., as described in the
evals/evals.jsonfile). - Boundary markers: There are no explicit instructions or delimiters defined to separate user input from the agent's logic, nor are there instructions for the agent to validate the security implications of user-requested configurations.
- Capability inventory: The skill enables the creation of a wide range of cloud resources across AWS, Azure, and GCP, including IAM roles, security groups, database instances, and storage buckets with potentially public access.
- Sanitization: The skill relies entirely on the underlying LLM's safety guardrails and the user's manual review of the generated Python code to detect and prevent the deployment of insecure infrastructure.
Audit Metadata