What happened
On Oct 6 ScienceLogic announced Skylar AI 2.7 (at Nexus Live 2026): Skylar Advisor gains an agent-routing trace and tool catalog showing how requests route and which internal/external MCP tools are considered; admins can disable specific agentic functions or MCP tools granularly; and a new upload gatekeeper inspects documents for exposed secrets, malicious instructions, and hidden-text manipulation before they enter the AI knowledge base, with quarantine and human-in-the-loop review.
Why it matters
RAG poisoning and prompt-injection via uploaded documents is a rising vector (hidden-instruction/hidden-text manipulation), and most guardrail products ignore knowledge-base ingestion. ScienceLogic adds inspection-at-ingestion plus per-MCP-function control in an established IT-ops platform — practical governance levers for enterprises granting AI agents operational authority in IT workflows.
Applicability
IT-operations and AIOps teams already on ScienceLogic (or evaluating service-centric observability) should assess Skylar AI 2.7 gatekeeper + MCP controls when giving Advisor access to ticketing/CMDB knowledge and tooling.