HKUST-KnowCompHKUST-KnowCompOfficialยท1 Agent Skills Included

DeepRefine-Skill

Refine knowledge graphs for better retrieval and Q&A

Refines Graphify knowledge graphs by analyzing past queries, diagnosing retrieval failures, and proposing edge and node fixes. Eliminates manual graph editing with dry-run reviews, evidence labels, and approval-gated writes that protect your data. Works inside Cursor, Claude Code, Codex, Gemini CLI, Copilot CLI, and OpenCode via one simple command.
npx skills add HKUST-KnowComp/DeepRefine-Skill --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install DeepRefine-Skill?โ–ผ

Run `npx skills add HKUST-KnowComp/DeepRefine-Skill --all -g -y` in your terminal to install the skill globally for your coding agent.

What does DeepRefine do to a knowledge graph?โ–ผ

It reviews your past queries, finds why retrieval failed, and proposes fixes like inserting edges, deleting wrong links, or renaming ambiguous nodes. Every change is reviewed with HIGH, MEDIUM, or LOW evidence labels before anything is written.

Is DeepRefine safe for my existing graph data?โ–ผ

Yes. It runs dry-run only by default, creates backups and checkpoints before any change, and never writes to your graph without your explicit approval.

Which coding agents work with DeepRefine?โ–ผ

It supports Cursor, Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, and OpenCode through a single /deeprefine command.

Do I need a Graphify knowledge base first?โ–ผ

Yes. DeepRefine refines an existing Graphify or LLM-Wiki knowledge graph, so you need to generate one with graphify before running refinement.

Related Repositories in Data & Analytics

View All in Data & Analyticsโ†’