Designing advanced RAG systems with agentic retrieval, GraphRAG, or evaluation-driven development.
Agentic RAG Architect
Agentic Loops • GraphRAG • Hybrid Search • RRF • RAGAS Evaluation
Best for
- ▸Design agentic RAG systems with plan-retrieve-grade-refine loops
- ▸Implement GraphRAG for global reasoning across large document sets
- ▸Build hybrid search pipelines combining vector, BM25, graph, and metadata
- ▸Set up evaluation-driven development with RAGAS metrics and golden datasets
What you'll get
- ▸Multi-hop retrieval pipeline design with query planning, relevance grading, and adaptive reformulation using supervisor/worker agent patterns
- ▸GraphRAG knowledge graph schema with entity extraction rules, community summaries, and global reasoning queries for corpus-level questions
- ▸Hybrid search architecture combining dense vectors, sparse BM25, and knowledge graph traversal with Reciprocal Rank Fusion scoring
- ▸RAGAS evaluation harness measuring faithfulness, answer relevancy, context precision, and context recall with automated regression detection
Document corpus description, query patterns, scale requirements, and current RAG pain points.
Complete agentic RAG architecture with retrieval strategy, orchestration pattern, evaluation plan, and phased implementation roadmap.
What's inside
“You are the Agentic RAG Architect — a senior AI systems architect specializing in next-generation Retrieval-Augmented Generation systems that go beyond naive retrieve-and-stuff patterns. You design RAG systems that treat retrieval as an agentic process: planning what to retrieve, grading relevance, ...”
Covers
Not designed for ↓
- ×Building basic RAG pipelines with simple retrieve-and-stuff patterns
- ×Vector database administration and operations tasks
SupaScore
88▼
Evidence Policy
Standard: no explicit evidence policy.
Research Foundation: 8 sources (1 paper, 2 official docs, 2 industry frameworks, 1 web, 2 academic)
This skill was developed through independent research and synthesis. SupaSkills is not affiliated with or endorsed by any cited author or organisation.
Version History
Initial release — agentic RAG with GraphRAG, hybrid search, and RAGAS evaluation
Works well with
© 2026 Kill The Dragon GmbH. This skill and its system prompt are protected by copyright. Unauthorised redistribution is prohibited. Terms of Service · Legal Notice