AI Security

Use this AI agent sandbox checklist to isolate files, network access, tool calls, runtime resources, and hostile inputs before production.

Use this AI agent rollback checklist to design previews, checkpoints, idempotency, approval gates, and recovery paths for production agent actions.

Use this AI agent secrets management checklist to keep tokens, keys, credentials, logs, memory, and retrieval indexes out of model-visible context.

Use this AI agent memory checklist to scope memory, validate writes, control retention, support deletion, and test memory poisoning risks.

Use this AI agent tool permissions checklist to scope tools, validate arguments, enforce authorization, require approvals, and log every tool call.

Use this AI agent data governance checklist to control access, retention, training use, memory, logs, deletion, export, and vendor data flows.

Use this AI agent incident response checklist to detect, contain, investigate, recover, and turn bad agent runs into stronger tests and controls.

Use this human-in-the-loop AI agent checklist to decide when risky tool calls need approval, how to review arguments, and how to log approval decisions.

A practical prompt injection testing checklist for AI agents covering direct attacks, indirect injection, malicious tool outputs, authorization, data leakage, and regression tests.

A practical AI agent security audit checklist covering prompt injection, RAG trust boundaries, tool permissions, guardrails, logs, cost, latency, and retesting.