The Agent Skills Directory

[Trusted Source Reference]: The skill exclusively references documentation and resources from official Cloudflare domains (e.g., developers.cloudflare.com), which is expected and secure given the skill's purpose.
[Indirect Prompt Injection Surface]: The skill defines patterns for ingesting external data, such as search parameters and JSON request bodies, which are then used to route to specific Durable Object instances or methods. This represents a potential surface for indirect prompt injection.
Ingestion points: Data enters through URL.searchParams and request.json() in the Worker handlers described in references/workers.md.
Boundary markers: The skill includes explicit instructions for input validation.
Capability inventory: Scripts perform Durable Object storage operations (SQL/KV), alarm management, and network responses.
Sanitization: The skill provides a clear pattern for sanitizing and validating input using the Zod library to ensure data conforms to expected schemas before processing.
[Secure Secrets Management]: The documentation correctly advises using platform-native secret management (e.g., wrangler secret put) rather than hardcoding credentials in configuration files like wrangler.jsonc or wrangler.toml.

durable-objects