AI Validator

Problem It Solves

Complex analysis, business advice, logical reasoning, and content generation often depend on data scope, time range, and unstated assumptions that users do not fully specify. When a model answers directly, small misunderstandings can be amplified into systematic errors. AI Validator turns “confirming understanding” into an explicit interaction protocol instead of relying on ad hoc behavior.

How It Works

The skill runs on a three-layer validation framework:
- Semantic understanding: it first restates the task goal, input scope, and key terms using phrasing like “Do you mean…”, aligning both sides on working memory.
- Contradiction detection: it scans internal consistency, factual consistency with public or trusted sources, and temporal or logical sequence; when conflicts appear, it pauses execution and asks for clarification.
- Assumption surfacing: before analysis, advice, or content generation, it lists key implicit assumptions, data limitations, risks, and boundary conditions.

It also provides trigger commands such as /validate, /check, /challenge, and /assumptions for full validation, quick confirmation, critical review, and assumption listing. The expected output style is professional, collaborative, cautious, and constructive, making it useful as a pre-check in data analysis, business advice, logical reasoning, and content generation.

Boundaries

This skill focuses on whether understanding is consistent; it does not replace verification against real data, domain judgment, or final business decisions. In high-risk contexts, it should be used to surface assumptions and clarify conflicts, not to generate final conclusions automatically. If the input lacks a trustworthy source, validation quality is still limited by context quality.