The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions and configuration examples reference numerous official packages and tools from trusted sources. These include NPM packages under the @cloudflare scope (e.g., @cloudflare/workers-types, @cloudflare/puppeteer, @cloudflare/playwright), official SDKs, and container images from Cloudflare's Docker Hub repository. These are documented as part of standard setup procedures for the platform.
[DYNAMIC_EXECUTION]: Documentation for Cloudflare Sandbox and Browser Rendering services naturally involves executing arbitrary logic (container execution or browser-side JavaScript). These capabilities are explicitly described as the primary purpose of the respective services and are implemented through official platform bindings.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building AI-powered applications, such as RAG (Retrieval-Augmented Generation) pipelines and autonomous agents. While these ingestion points process unstructured user input, the skill emphasizes safety by using parameterized SQL queries in its database integration examples to prevent secondary injection vulnerabilities.
[CREDENTIALS_UNSAFE]: The documentation includes numerous placeholders for sensitive values like API_KEY, CLOUDFLARE_API_TOKEN, and OPENAI_API_KEY. The skill correctly instructs users to manage these values using secure platform features such as wrangler secret put, rather than hardcoding them in configuration files.

cloudflare