The Agent Skills Directory

Prompt Injection (SAFE): No instructions designed to override agent behavior, bypass safety protocols, or reveal system prompts were found. The language is purely instructional and technical.
Data Exposure & Exfiltration (SAFE): There are no hardcoded credentials, API keys, or access to sensitive local file paths (e.g., .ssh, .aws). The network operations in the examples are standard for Cloudflare Workers (handling HTTP requests via Hono).
Obfuscation (SAFE): No Base64, zero-width characters, or homoglyphs were detected. The content is entirely human-readable markdown and TypeScript.
Unverifiable Dependencies & Remote Code Execution (SAFE): The skill references 'wrangler' and 'cloudflare-worker-base', which are standard, trusted tools for the Cloudflare ecosystem. No remote script execution (e.g., curl|bash) is present.
Indirect Prompt Injection (LOW): The skill provides patterns for processing external data from HTTP requests and message queues.
Ingestion points: README.md contains examples using c.req.json() and processing batch.messages.
Boundary markers: None present in the example code.
Capability inventory: No executable scripts provided; the logic is confined to documentation examples.
Sanitization: No explicit sanitization or validation of the message body is shown in the quick examples, which is a common surface for indirect injection if the processed content is later used in an LLM prompt.

cloudflare-queues