About this sample
KB AI Review Lite performs an AI-driven accuracy validation pass on knowledge base documents stored in a SharePoint document library. The skill reads each document, extracts concrete factual claims such as URLs, version numbers, ports, timeouts, and procedural steps, and validates them against category-specific sources including Microsoft MSRC, vendor documentation, and public runbooks.
On first run the skill creates all 9 required review columns automatically. A one-time manual step is required to enable unlimited text length on two columns — the skill detects this condition, stops before writing any data, and provides exact step-by-step instructions. All subsequent runs proceed directly to the review with no interruption.
Review output is written entirely as library metadata. The skill never modifies the original document content. Each reviewed file receives an AI Review Summary, a structured Markdown draft payload with findings and evidence, a Review Score displayed as a percentage bar, a color-coded AI Review State pill, an inferred KB Category, and a Next Review Date calculated from a category-based cadence.