azure-advisor-cost-optimization

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill processes data from external Azure Advisor APIs, which presents a surface for potential indirect prompt injection.
  • Ingestion points: Data enters the agent's context through tools such as advisor_recommendation_list and advisor_recommendation_summary as specified in references/tools-and-recovery.md.
  • Boundary markers: To mitigate risks, the skill uses structured Markdown tables and specific headings (defined in references/table-format.md) to clearly delineate external data from instructional content.
  • Capability inventory: The skill has the ability to update recommendation states via advisor_recommendation_update (references/state-updates.md) and provide remediation guidance through handoffs (references/selection-and-handoff.md).
  • Sanitization: The skill implements sanitization by URL-encoding resource identifiers when constructing portal links as seen in references/table-format.md.
  • Command Execution: In references/scope-resolution.md, the skill identifies a fallback mechanism using az account show and az account list. This command execution is restricted to resolving subscription scope and is a standard procedure for cloud administrators to verify their environment context.
  • Safe Handling of Sensitive Operations: The skill is designed with a strong emphasis on read-only operations. As described in references/selection-and-handoff.md, it offers a prevalidation flow that provides safety checks without authorizing deletion. For state changes, the skill relies on a batch update process that requires server-side approval, ensuring that no resources are modified without explicit oversight.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 03:11 AM