guardduty-cost-optimization
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFE
Full Analysis
- Managed Data Ingestion Surface: The skill processes data from Amazon GuardDuty and CloudWatch metrics. While this represents a surface area for potential indirect prompt injection, the skill incorporates comprehensive mitigations by instructing the agent to treat all ingested identifiers and statistics as inert data. It specifically warns against following any instruction-like strings found in the environment.
- Least Privilege Architecture: The skill adheres to the principle of least privilege, relying exclusively on read-only tools to collect inventory and usage data. It contains no instructions to perform mutation operations, ensuring that any optimization actions require human intervention and review.
- Security-Centric Recommendations: The skill prioritizes security posture by requiring that any plan reduction be justified by both cost and signal (finding activity). This prevents cost optimization from inadvertently creating security gaps, as all recommendations are presented as cost-vs-risk tradeoffs.
Audit Metadata