The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill guides users to install necessary development dependencies from the npm registry, including wrangler, @cloudflare/workers-types, and @cloudflare/vitest-pool-workers. These are official packages maintained by Cloudflare, a well-known service provider.
[COMMAND_EXECUTION]: The skill uses the Wrangler CLI (npx wrangler) for local development, state management, and deployment. These commands are standard for the Cloudflare Workers ecosystem and are used as intended for service configuration and project initialization.
[DATA_EXFILTRATION]: The skill explicitly warns users against storing sensitive information in configuration files. In SKILL.md under 'Critical Prohibitions', it instructs users to use wrangler secret put for secrets management instead of including them in wrangler.toml, which is a recognized security best practice.
[PROMPT_INJECTION]: As a developer tool for building web services, the skill naturally defines attack surfaces for indirect prompt injection (Category 8).
Ingestion points: The fetch, queue, and email handlers documented in references/runtime.md ingest untrusted data from HTTP requests, message queues, and inbound emails.
Boundary markers: The provided code examples are generic templates and do not include specific delimiters or 'ignore' instructions for data processed by downstream AI components.
Capability inventory: The skill documents capabilities to interact with external networks (Fetch API), persistent storage (KV, R2, D1), and AI inference (Workers AI), as detailed in references/bindings.md.
Sanitization: The documentation focuses on functional implementation; developers are responsible for implementing data validation and sanitization for their specific use cases.

cloudflare-workers