The Agent Skills Directory

[SAFE]: The skill provides comprehensive technical references for Cloudflare services. All detected command-line examples and code snippets use official Cloudflare tools (e.g., Wrangler, cloudflared) or well-known community tools (e.g., Pulumi, Terraform, Hono).
[EXTERNAL_DOWNLOADS]: The skill references downloads for official tools like wrangler from npm and cloudflared via brew. It also includes an example in the Sandbox reference for installing code-server from a well-known domain. These are standard practices for the documented use cases.
[COMMAND_EXECUTION]: Various shell commands are provided for resource management (e.g., wrangler deploy, cloudflared tunnel create). These are legitimate management commands for the Cloudflare platform.
[PROMPT_INJECTION]: Scanned all 257 files for prompt injection markers. No patterns attempting to bypass agent safety or override system instructions were detected. Occurrences of 'CRITICAL' or 'IMPORTANT' in the text are used for technical constraints (e.g., 'CRITICAL: proxyToSandbox MUST be called first').
[CREDENTIALS_UNSAFE]: No hardcoded credentials were found. Documentation correctly instructs users to use wrangler secret put or .env files for secret management. Placeholders like <token> or YOUR_API_TOKEN are used throughout.

cloudflare