microsoft-docs — agentic threat model

Not certain from the listing — likely uses Azure OpenAI or similar Microsoft-hosted foundation models. Primary threats include prompt injection to bypass retrieval constraints or model reprogramming to output malicious instructions disguised as official documentation.

Queries official Microsoft documentation (Azure, .NET, M365, etc.). The primary threat is documentation-retrieval poisoning if an attacker can manipulate the source documentation or if the vector database/search index is compromised, leading to the delivery of malicious configuration advice.

Not certain from the listing — likely built on Semantic Kernel or AutoGen given the Microsoft ecosystem. Threats include insecure tool integration with the documentation search API and potential prompt injection via search results (indirect prompt injection).

Not certain from the listing — presumably hosted within Microsoft's cloud infrastructure (Azure). Threats include unauthorized access to the search API endpoints and potential SSRF if the document retrieval mechanism can be coerced into querying arbitrary internal URLs.

Not certain from the listing — no specific evaluation, logging, or guardrail mechanisms are detailed in the public directory listing. Gaps here could lead to undetected drift or silent failures in document retrieval accuracy.

Not certain from the listing — while it is an official Microsoft skill, specific compliance certifications (like SOC2, ISO) or identity/authorization controls for this specific agent are not detailed in the listing.

Designed as an 'Agent Skill' or utility agent. It is highly likely to be integrated into larger multi-agent systems or orchestrators to provide technical context. The primary threat is cascading trust abuse, where a compromised orchestrator uses this agent to retrieve misleading configuration data.