The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes boilerplate code for handling external data, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters via request.url, request.body, and request.headers in the fetch handlers within src/index.ts and src/counter.ts.
Boundary markers: None present, as the examples are standard web worker implementations.
Capability inventory: The worker scripts possess write capabilities to Cloudflare KV (env.KV.put), D1 Database (env.DB.prepare().run()), and R2 Storage (env.BUCKET.put).
Sanitization: The skill demonstrates safe practices by using parameterized SQL queries (.bind()) and the URL API for parsing, which mitigates common injection vectors.
[SAFE]: The skill promotes several security best practices:
Supply Chain Security: Recommends using trustedDependencies to block post-install scripts and the socket CLI for auditing third-party packages.
Credential Management: Explicitly warns against logging secrets (env.API_KEY) and provides instructions for managing them via wrangler secret put.
Official Tooling: Recommends using official Cloudflare scaffolding tools (create-cloudflare) and maintains current versions of core developer dependencies.

bun-cloudflare-workers