# Red-team scenario library run brief

Use this brief to run a controlled red-team session before an AI workflow receives sensitive data, broad access, or tool authority. The session should produce evidence: scenario inputs, expected behavior, actual behavior, severity, mitigation owner, and a regression test that prevents the same failure from returning.

## Run sequence

- Confirm scope, model version, tools, data sources, and logging.
- Execute representative attacks across retrieval, chat, memory, and tools.
- Record pass or fail with severity and evidence.
- Convert failures into release gates, mitigations, or accepted risks.