# Scenario Catalog

## 01 - AI Prompt Injection

Shows an AI agent attempting to generate overly broad SQL. Oracle Deep Data Security limits the result based on the end-user context.

## 02 - Shared App Account

Shows the risk of a technical application account shared by multiple users. The relevant control becomes the end-user context, not only the connection pool account.

## 03 - PII Row/Column/Cell

Shows row, column, and cell-level access control. An employee sees their own record, a manager sees the team with SSN hidden, and HR sees sensitive attributes.

## 04 - View Bypass / MAC

Shows how a view can become an alternate access path and how `USE DATA GRANTS ONLY` enforces the base table policy.

## 05 - Legacy App AI Extension

Shows a legacy application extended with an AI agent without rewriting all authorization logic. The agent accesses the same dataset, but Oracle Deep Data Security limits rows and columns by end-user context.

## 06 - RAG Vector Classified Docs

Shows RAG/vector search with classified documents. Retrieval may attempt to fetch all chunks, but the database returns only what the classification and persona allow.

## 07 - Audit Evidence With Data Safe

Shows how to generate evidence with Unified Audit and how to position OCI Data Safe for activity auditing, reporting, and sensitive data access validation.

## Suggested Next Scenarios

- BI by region, branch, and cost center.
- Controlled write operations with `UPDATE`, `INSERT`, and `DELETE`.
- Database Vault complement to block DBA access.
- AVDF/SIEM integration for on-premises, Exadata, or heterogeneous environments.