building-ai-agent-on-cloudflare
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: References official Cloudflare starter templates and reference implementations from the Cloudflare organization on GitHub.
- [COMMAND_EXECUTION]: Provides instructions for using official Cloudflare development tools, including the Wrangler CLI and project initialization commands, which are standard for the documented platform.
- [INDIRECT_PROMPT_INJECTION]: The skill documents architectural patterns where agents ingest data from WebSockets, tool outputs, and vector databases (ingestion points). These examples demonstrate capabilities like AI model invocation, SQL database access, and network operations across multiple files (SKILL.md, references/agent-patterns.md). While providing a robust framework for development, these patterns demonstrate an attack surface where external data is processed by an LLM without explicit sanitization or strict boundary markers in the provided snippets.
Audit Metadata