What problem does it solve? CrewAI agents execute tools and process inbound messages without built-in security controls, leaving them exposed to prompt injection, PII leakage, and uncontrolled tool usage. This Skill adds Arcjet Guard policy enforcement to CrewAI workflows so risky tool calls are rate-limited and malicious inputs are blocked before execution. ## Core Features & Use Cases - Process-wide tool gating: Register PRE_TOOL_CALL hooks with register_arcjet_hooks and ToolPolicy to enforce rate limits and security rules on every crew-executed tool. - Standalone tool wrapping: Use guard_tool to protect a CrewAI BaseTool you invoke directly, raising ArcjetDeniedError on policy denial. - Inbound message screening: Screen user messages for prompt injection with core guard before crew.kickoff. - Use Case: A support agent crew looks up orders via a tool. Apply a TokenBucket rate limit of 10 lookups per minute per authenticated user, and block prompt injection in inbound messages before the crew starts. ## Quick Start Ask your AI agent to integrate Arcjet Guard into your CrewAI project by registering PRE_TOOL_CALL hooks with rate limits on your tools and screening inbound messages for prompt injection before crew kickoff.
