cost-analysis
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- API Data Ingestion Surface: The skill ingests data from Azure Cost Management and Resource Graph APIs to generate spend reports. While this creates a surface for indirect prompt injection from external API responses, the skill mitigates this by instructing the agent to differentiate between measured costs and inferred causes, use specific evidence labels (e.g., 'Actual cost', 'Estimate'), and disclose when results are partial or incomplete.
- Controlled Fallback Execution: The skill permits the use of official Azure CLI and PowerShell commands if primary MCP tools are unavailable. This potential command execution is safely managed by requiring the use of authenticated vendor clients, prohibiting local script interpreters for data parsing, and mandating explicit user confirmation for any configuration changes or write operations.
- Resource and Cost Guardrails: Operational risks are minimized through built-in requirements for manual confirmation before the agent performs cost-impacting actions, such as enabling cost-analysis add-ons, changing service tiers, or creating budgets, ensuring all changes align with user intent and least-privilege principles.
Audit Metadata