dd-aws-integration
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches auxiliary skill instructions from the official Datadog-labs GitHub repository. The implementation includes shell-based validation of the skill name to prevent directory traversal or command injection.
- [COMMAND_EXECUTION]: Utilizes
terraform,curl, andawsCLI for infrastructure provisioning and state verification. Secret handling for these commands is performed via stdin redirection, avoiding credential exposure in shell history or process metadata. - [INDIRECT_PROMPT_INJECTION]: Supports loading external instructions from other skills within the vendor's repository. The skill explicitly instructs the agent that these external files are for reference only and cannot override existing safety protocols.
- Ingestion points: SKILL.md files downloaded from github.com/datadog-labs/agent-skills.
- Boundary markers: Present. Instructions explicitly state that external skills cannot override user or safety rules.
- Capability inventory: Subprocess calls (terraform, aws, curl) and file reading/writing operations.
- Sanitization: Input for skill names is validated against a strict regex pattern using a shell case statement.
Audit Metadata