superragskills

The complete agent skill pack for designing, building, evaluating, and deploying retrieval-augmented generation applications.

superragskills helps coding agents guide a user from "I want a RAG app" to a production-ready plan and implementation. It covers basic RAG, hybrid RAG, reranked RAG, metadata-filtered RAG, hierarchical RAG, GraphRAG, agentic RAG, corrective/self-RAG, multimodal RAG, structured-data RAG, local/offline RAG, and cloud deployments on AWS and Azure.

What It Includes

• A primary agent skill: skills/superrag-build/SKILL.md
• RAG type decision matrix
• Full discovery questionnaire
• From-scratch build playbook
• Provider and stack matrix
• AWS, Azure, local server, Docker, and Kubernetes deployment playbooks
• Evaluation and observability guide
• Security and governance guide
• Output templates and examples for coding agents

Layout

superRAG-build/
  skills/
    superrag-build/
      SKILL.md
      agents/openai.yaml
      references/
      examples/
  .claude-plugin/
  .cursor/rules/
  AGENTS.md
  CLAUDE.md

Install Manually

Claude Code:

cp -r skills/superrag-build ~/.claude/skills/superrag-build

Codex:

mkdir -p .agents/skills
cp -r skills/superrag-build .agents/skills/superrag-build

Cursor:

mkdir -p .cursor/skills
cp -r skills/superrag-build .cursor/skills/superrag-build

For Cursor rule-based fallback, keep .cursor/rules/superrag-build.mdc in the repo.

Example Prompts

Use superRAG-build. Interview me and decide which RAG architecture I need.

Use superRAG-build to design and implement a production RAG app for PDFs, tables, and internal tickets, deployable on Azure.

Use superRAG-build to compare AWS Bedrock Knowledge Bases, Azure AI Search, and a self-hosted Qdrant stack for my requirements.

Star History

Privacy

This repository is built from public documentation, research, and generalized engineering lessons. It should not include private application code, private customer data, internal file paths, or proprietary corpus details.