Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/nvidia-ai-blueprints/rag/agents-mdgit clone --depth 1 https://github.com/NVIDIA-AI-Blueprints/ragWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/nvidia-ai-blueprints/rag/agents-md)<a href="https://agentmods.dev/instructions/nvidia-ai-blueprints/rag/agents-md"><img src="https://agentmods.dev/badge/instructions/nvidia-ai-blueprints/rag/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00951 | $0.00951 |
| Opus 5 | $0.00476 | $0.00476 |
| Sonnet 5 | $0.00190 | $0.00190 |
| Haiku 4.5 | $0.00095 | $0.00095 |
Grade A, and why
rag AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
NVIDIA RAG Blueprint
Reference implementation for a Retrieval Augmented Generation pipeline. Python 3.11+ backend (FastAPI + LangChain), React/TypeScript frontend, deployable via Docker Compose or Helm.
Project structure
src/nvidia_rag/
├── rag_server/ # RAG query/response server (FastAPI)
├── ingestor_server/ # Document ingestion server (FastAPI)
└── utils/ # Shared utilities
frontend/ # React + TypeScript UI (pnpm)
deploy/
├── compose/ # Docker Compose files and env configs
└── helm/ # Helm charts (standard + MIG-slicing)
docs/ # User-facing documentation (Sphinx, RST/MD)
tests/
├── unit/ # No network calls allowed
└── integration/ # Network calls permitted
notebooks/ # Jupyter notebooks for evaluation and examples
Development commands
Backend (Python)
uv sync # Install all deps
# Optional: RAGAS benchmark CLI (see scripts/eval/README.md)
# uv sync --project scripts/eval
uv run pytest tests/unit/ # Unit tests
uv run pytest tests/integration/ # Integration tests
ruff check --fix src/ # Lint + autofix
ruff format src/ # Format
pre-commit run --all-files # Run all pre-commit hooks
Frontend (TypeScript)
cd frontend
pnpm install
pnpm run dev # Dev server
pnpm run lint # ESLint
pnpm exec tsc --noEmit # Type check
pnpm run test:run # Tests
Code conventions
- Python: Ruff for linting and formatting (line-length 88, double quotes, space indent). Config in
pyproject.toml. - Type hints: Required on all function signatures.
- Imports: Sorted by isort via Ruff. No in-function imports.
- Tests: Mirror source tree (
src/nvidia_rag/rag_server/server.py→tests/unit/rag_server/test_server.py). - Frontend: ESLint + TypeScript strict mode. Function components with hooks.
- Env files:
deploy/compose/nvdev.env(NVIDIA-hosted NIMs) anddeploy/compose/.env(self-hosted). These are the source of truth for Docker deployments — shell-only exports are lost on restart.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 89 lines · 951 tokens per session scan A d0eed583831b
rag AGENTS.md is an instructions file published in the GitHub repository NVIDIA-AI-Blueprints/rag (754 stars, last pushed yesterday), licensed Apache-2.0. It adds 951 tokens to every session, about $0.0048 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
RAGLight CLAUDE.md
Instructions for Bessouat40/RAGLight, covering claude.md, commands, install dependencies, run all tests and run a single test module.
ragflow CLAUDE.md
Claude Code instructions for infiniflow/ragflow, a project described as: RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs.
gemini-cli-extension GEMINI.md
Instructions for pinecone-io/gemini-cli-extension, covering pinecone extension for gemini cli, available agent skills, key concepts & setup and available mcp tools.
FileSearchStore-extension GEMINI.md
Instructions for tanaikech/FileSearchStore-extension: You are an expert in managing File Search Stores for Gemini API using all tools of the MCP server "file-search-store-extension".
GPT-RAG release.instructions.md
Instructions for Azure/GPT-RAG, a project described as: Enterprise-grade accelerator for agentic RAG on Azure. Built on Microsoft Foundry with Foundry IQ as the default retrieval backend, Microsoft Agent Framework orchestration, Zero-Trust architecture and IaC.
the-architect CLAUDE.md
Instructions for Hainrixz/the-architect, covering the architect, non-negotiable rules — these apply on every turn, forever, state machine, discovery and deep dive.