evidence-ontology

To ground LLM reasoning with traceable evidence, you decompose prompts into canonical ontology elements and tag each claim using a 13-tag Evidence Ontology. This standardizes reasoning and ensures all derivations are traceable to a single set of rules.

An evidence ontology is a canonical set of 13 tag types used to decompose prompts into baseline, ideal, and delta buckets. It works by tagging each claim to surface gaps, prioritize questions and hypotheses, and guide design and documentation decisions.

You validate artifacts and prevent context rot by running mid-session re-grounding and using the verifying-evidence workflow. This checks loaded claims against the canonical ontology rules to reduce drift and maintain artifact coherence.

Yes, the evidence tagging approach supports frontmatter metadata validation. It provides explicit tag formatting rules and validation guidance to keep documentation artifacts coherent across multiple sessions.

You apply decision buckets by separating context into baseline, ideal, and delta categories. This decomposition surfaces gaps between current and ideal states, allowing you to prioritize questions and hypotheses for design decisions.

The 13-tag ontology requires loading the complete rule set into context at kickoff, which consumes context window space. Mid-session re-grounding is necessary to prevent context rot and maintain artifact accuracy during long tasks.