AI agent and Inference Hooks policies

Where the AI agent and Claude Inference Hooks policy templates live, next to the endpoint recommendations.

Recommendations 1 through 9 on Recommended Policy Configurations cover endpoint triggers: browser uploads, clipboard, desktop apps, cloud sync, git push, removable media, and print. Those triggers do not fire inside Claude Code, Cursor, VS Code, or Claude Inference Hooks.

Create the agent policies separately, under Configuration > Policies > Exfiltration > AI Agent Security. The templates, hook types, and actions are on Hook policy recommendations. The connect flow is on Connect Inference Hooks.

If you already use	Also create
Recommendation 1, AI chat uploads and pastes	Inference Hooks, User Prompts, content scanning on, for claude.ai and Cowork. Device hooks for Claude Code, Cursor, and VS Code
Recommendation 3, credentials	Device hooks: User Prompts, Tool Calls, and Shell Commands, Block, content scanning on
Recommendation 4, source code	Device hooks, User Prompts and Tool Calls, Block, after a monitor week. Inference Hooks, User Prompts, shadow then Block
Recommendation 5, PHI	Inference Hooks, User Prompts, for Claude Enterprise chat, plus device hooks with PHI detectors
Recommendations 7 and 8, PCI and financial files	Inference Hooks, User Prompts, shadow then Block

On an AI Agent Security policy, content scanning off is enforced. It does not wait for a detector. Block denies every event the policy covers. Monitor records a violation on every event and lets it through. Turn content scanning on, and add detection rules, unless you intend to stop or record all of that traffic. This matches other endpoint exfiltration policies.