# AI RAG Defense

> Token reduction, retrieval fidelity, and prompt-injection defense layers for the Context Engine. Covers a Summarizer agent for proactive context reduction, micro-context engineering (prompt design inside agents), high-fidelity RAG with source-metadata citations, layered untrusted-context defenses against prompt injection, and grounded-reasoning validation that prevents hallucination. Use this skill when: - Reducing token costs by adding a Summarizer agent - Designing system prompts inside specialist agents (micro-context) - Adding citations and source metadata to RAG outputs - Defending agents against prompt injection / data poisoning - Validating agent reliability across high-fidelity / summarization / grounded test cases - Preventing hallucination when retrieval returns no result

- Skill: `ebarti/ai-rag-defense` (Agent Skill, multi-file: 25 files)
- Install (CLI): `npx skillmds@latest add ebarti/ai-rag-defense`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ebarti/ai-rag-defense/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ebarti (https://skillmd.com/u/ebarti)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/ebarti/ai-rag-defense

---


# AI RAG Defense

Knowledge from "Context Engineering for Multi-Agent Systems" (Chapters 6-7). The reduction, fidelity, and defense layers that turn a working engine into a trustworthy one.

## Quick Start

1. Check `guidelines.md` to find which files to load
2. Load only relevant files (each topic has knowledge.md, rules.md, examples.md)
3. Apply guidance to your work

## Contents

### References

| Category | Purpose |
|----------|---------|
| `summarizer-agent` | Cost-management agent for proactive context reduction |
| `micro-context-engineering` | Prompt design inside agents (macro vs micro) |
| `high-fidelity-rag` | Source-metadata-on-chunks pattern, citations-mandatory Researcher |
| `input-sanitization` | Untrusted context boundary, prompt-injection threat model + smells |
| `grounded-reasoning` | Report-negative-finding pattern, multi-case validation |

### Workflows

| Workflow | Purpose |
|----------|---------|
| `workflows/add-summarizer.md` | Integrate Summarizer agent for proactive context reduction |
| `workflows/add-citations.md` | Upgrade RAG to high-fidelity (source metadata + cite) |
| `workflows/defend-against-injection.md` | Add untrusted-context boundary + run grounded-reasoning validation suite |

## Guidelines

See `guidelines.md` for task-based file selection.