The Agent Skills Directory

Indirect Prompt Injection Surface: The skill provides extensive documentation for building AI-powered agents using the Cloudflare Agents SDK. This architecture involves ingesting untrusted data through various handlers (e.g., onChatMessage, onMessage, onEmail) and provides the agent with capabilities such as SQL access, network operations, and scheduled tasks. While these are core features of the platform, the documentation includes guidance on using parameterized queries to mitigate data-related risks. Developers using these patterns should implement clear delimiters and instructions for the AI to ignore potentially malicious content within processed data.
Remote Script Execution Examples: Several reference files include examples of downloading and executing scripts from remote sources, such as the installation of cloudflared for Tunnels or code-server within the Sandbox environment. These examples are typical for the integration and management of the documented services and point to well-established domains.
Platform-Native Security Patterns: The documentation consistently emphasizes the use of platform-native security features. This includes recommending the use of environment secrets via wrangler secret put instead of hardcoding API tokens, and mandates the use of prepared statements with the D1 database to prevent SQL injection vulnerabilities.

cloudflare