Run Cypher probes and LLM evaluation to produce a Markdown quality report for Neo4j knowledge graphs.
npx skills add https://github.com/YH-05/quants --skill kg-quality-check
Please help me install this Agent Skill. Skill: kg-quality-check Source: https://github.com/YH-05/quants/tree/main/.claude/skills/kg-quality-check Command: npx skills add https://github.com/YH-05/quants --skill kg-quality-check
This skill measures and reports the data quality of a knowledge graph (KG) by running seven quantitative probes on a Neo4j KG (v2.2) and applying an LLM-based judge to evaluate claim/fact accuracy, discovery potential, and graph integrity. It outputs a Markdown report with scores, issues, and concrete improvement suggestions.
Run the kg-quality-check workflow to generate a comprehensive quality report for the current KG.
High-intent search queries and answers about installing and using this skill.
Knowledge graph quality is measured by running seven quantitative Cypher probes against the Neo4j schema and applying an LLM-based judge to evaluate claim accuracy, discovery potential, and graph integrity, yielding a Markdown quality report.
Yes, knowledge graph quality checks use read-only Cypher queries via mcp__neo4j-cypher__read_neo4j_cypher, meaning no writes are performed and the Neo4j database remains completely unmodified.
LLM evaluation assesses knowledge graph data quality by judging the precision of claims and facts, hypothesizing potential discoveries, and generating a structured critique with concrete improvement suggestions across seven quantitative categories.
To generate a knowledge graph quality report after data ingestion, run the kg-quality-check workflow to execute Cypher probes and LLM-based evaluation, producing a Markdown document with scores, issues, and improvement suggestions.
The knowledge graph quality check applies to Neo4j KG v2.2 with 14 nodes and 31 relationships, running seven quantitative probes tailored to this specific schema configuration.
A knowledge graph quality report evaluates seven quantitative categories computed via Cypher probes, covering fact accuracy, discovery potential, and graph integrity to produce scores and issues for review.
YH-05Ingest graph-queue JSON into Neo4j with idempotent MERGE-based insertion.
ShutovKSGenerate GRACE project health reports from artifacts, codebase metrics, and verification status.
JasonWarrenUKAnalyze user queries to produce Cypher patterns for Neo4j graph databases.
labs21-devProfile datasets to reveal structure, data quality, and relationships.
