AgentSkillsCN

cloudflare-workers

实施并配置 Cloudflare Workers、D1、R2、KV 以及 Wrangler。当您需要构建或配置 Workers 应用、D1 数据库、R2 存储桶、KV 命名空间、wrangler.toml 文件、绑定关系、Wrangler 开发环境、D1 迁移、.dev.vars，或通过 Wrangler 进行部署时，可使用此功能。在使用当前 API 时，优先参考 Cloudflare 官方文档以及 MCP/Context7 的相关指南。

SKILL.md
--- frontmatter
name: cloudflare-workers
description: Implements and configures Cloudflare Workers, D1, R2, KV, and Wrangler. Use when building or configuring Workers apps, D1 databases, R2 buckets, KV namespaces, wrangler.toml, bindings, wrangler dev, D1 migrations, .dev.vars, or deploying with Wrangler. Prefer Cloudflare docs and MCP/Context7 for current APIs.

Cloudflare Workers Skill

Quick Start

Use this skill when the user is implementing or configuring Cloudflare Workers, D1, R2, KV, Wrangler, or bindings. Terraform is out of scope until explicitly introduced as a feature; see AGENTS.md.

Scope

  • •In scope: Worker code (TypeScript), wrangler.toml, bindings (D1, R2, KV), Wrangler CLI (dev, deploy), secrets (.dev.vars), D1 migrations, R2 S3-compatible API, KV get/put. Worker and Docker are 1:1 (same app, same config).
  • •Out of scope: Terraform (ignore infra/terraform/ unless the user explicitly asks for it); Discord/SimHub (use their skills).

Conventions (this repo)

  • •Secrets: Keep in .dev.vars (or env-specific files). Never commit secrets; .dev.vars is in .gitignore.
  • •Bindings: Wire D1/R2/KV in wrangler.toml; names/IDs from Cloudflare Dashboard or your own automation. Same config for Docker and wrangler dev.
  • •CPU: Workers have a 50ms CPU limit per request; design handlers to stay under or offload to Queues/Cron.
  • •Stack: Workers (TypeScript), D1 (SQLite), R2 (S3-compatible), KV (caching). See ADR-001 and cost-optimized ADR.

Key Resources

  • •Wrangler: Local dev wrangler dev; deploy wrangler deploy. Use wrangler d1 execute for migrations.
  • •D1: SQLite at the edge; use prepared statements and migrations. Free tier: 5M reads/day, 100K writes/day.
  • •R2: S3-compatible; use AWS SDK or R2 API for plugin binaries and static assets. Zero egress fees.
  • •KV: Key-value cache; use for hot data to reduce D1 reads (e.g. session or config cache).

Documentation

  • •Prefer Cloudflare official docs and Cloudflare MCP (e.g. search_cloudflare_documentation) or Context7 for current API and examples.
  • •ContextStream (when available): Before Grep/Read in apps/api/, use ContextStream search (mode=hybrid) for "Worker", "D1", "auth", "wrangler". After Worker/ADR decisions, capture in ContextStream (event_type=decision) with path to wrangler.toml or ADR. Use graph(action="impact", target="...") before refactoring Worker code. See development guide.
  • •Project docs: infrastructure, database schema, API. API docs: Each Worker endpoint is documented in docs/api/openapi.yaml and served in Swagger at /api-docs (local and Docker). When adding or changing routes, update the OpenAPI spec; smoke test ensures API documentation loads.

Checklist (when adding or changing a Worker)

  • • wrangler.toml defines name, compatibility date, and bindings (d1_databases, r2_buckets, kv_namespaces).
  • • Secrets are in .dev.vars and not committed.
  • • D1 migrations (if any) live in the Worker app and are run via wrangler d1 execute or CI.
  • • Bindings in wrangler.toml match your D1/R2/KV resources (Dashboard or automation); Worker and Docker use the same config.