devops-finops

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous command-line examples for cloud provider CLIs (AWS, Azure, GCP) and Kubernetes (kubectl, helm). These commands are intended to query cost data, manage budgets, and perform resource optimization tasks such as purchasing reservations and stopping instances.
  • [EXTERNAL_DOWNLOADS]: The skill documents the installation of Kubecost using Helm (helm repo add kubecost https://kubecost.github.io/cost-analyzer/). This reference targets a well-known service and is documented neutrally as part of the Kubernetes cost optimization workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from cloud environment resource metadata (tags, resource names) and external tool outputs (Kubecost, cloud recommendations). It possesses capabilities to modify infrastructure and financial commitments, creating a surface for indirect prompt injection if resource names contain malicious instructions.
  • Ingestion points: Resource tags and names retrieved via boto3 (SKILL.md), AWS Tagging API, and GCP Billing Export (references/finops-practices.md).
  • Boundary markers: None identified in the provided templates.
  • Capability inventory: Resource modification via ec2.stop_instances (SKILL.md) and financial commitment via aws savingsplans purchase-savings-plan (SKILL.md).
  • Sanitization: None identified in the provided logic snippets.
  • [SAFE]: All identified Python and CLI snippets use standard libraries and tools (boto3, aws-cli, gcloud, kubectl) for their intended administrative purposes. Resource IDs and ARNs used in examples are generic placeholders.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 06:34 PM