Source: cloudflare/skills · Maintainer: Cloudflare · License: Apache-2.0 · Checked: 24 October 2026
What is the Cloudflare skill?
The cloudflare skill is Cloudflare's official Agent Skill for choosing Cloudflare products. You describe what you want to build, and it helps your coding agent pick the products that fit, pick the right CLI, and load the matching skill or current docs before it writes any code.
Its SKILL.md description reads: "Discover and choose Cloudflare products for apps, APIs, AI agents, storage, networking, and security. Use for architecture and product selection, including when the user describes a need without naming a Cloudflare product; then find the relevant skill or documentation."
That last part is the point. As the skill puts it, "A user asking for uploads, background jobs, or document search may not know to ask for R2, Queues, Workflows, or AI Search." It is the entry point to the rest of the cloudflare/skills repo, and the most installed skill in it: its skills.sh page listed 135.8K installs (snapshot, 24 October 2026).
What does it do?
The skill is a routing guide. It holds very little product detail itself and instead tells the agent where to look.
It picks the CLI first. The rule is short:
| Project | CLI the skill chooses |
|---|---|
Has wrangler.jsonc, wrangler.json or wrangler.toml, and no cloudflare.config.ts |
Wrangler, plus the wrangler skill |
| Anything else, including a new project | The Cloudflare CLI (cf), plus the cf-cli skill if installed |
In the second case it says: "Do not install Wrangler or create a Wrangler configuration file." When guidance names a Wrangler command, the agent should find the cf equivalent with cf cli search "<task>" or Cloudflare's Wrangler to cf reference "instead of translating from memory." Cloudflare's Use cf with coding agents page covers that side.
It maps needs to products. The core is a table of more than 90 rows titled "What are you trying to build?" Each row pairs a task with a product, when to choose it, and a skill or reference. A few examples from the SKILL.md:
- Store uploads, downloads or large objects: R2, paired with D1 "when searchable metadata needs SQL."
- Run a job that retries, waits and resumes across steps: Workflows.
- Coordinate chat rooms, games or bookings: Durable Objects and the
durable-objectsskill. - Connect to an existing PostgreSQL or MySQL database: Hyperdrive.
- Protect forms from automated abuse: Turnstile and the
turnstile-spinskill. - Send transactional email: Email Service and the
cloudflare-email-serviceskill. - Require employee login before an internal app: Access and the
cloudflare-oneskill.
The rows cover apps and APIs, storage, AI (Workers AI, AI Search, Vectorize, AI Gateway, the Agents SDK), networking (DNS, Tunnel, Load Balancing), security (WAF, Bot Management, API Shield), media, email, observability and infrastructure as code.
It makes a default call on hosting. "Recommend Workers and Workers Static Assets for new websites and applications, including static sites, SPAs, and full-stack apps." It adds that Workers "is recommended for all new projects" and that existing Pages deployments should be preserved during unrelated maintenance. See Cloudflare's Workers Static Assets docs.
It has a caching rule. It prefers Workers Cache and says to choose the Cache API or KV caching "only when a concrete requirement cannot be met by Workers Cache."
It ships references. The folder has 27 reference topics under references/, each with a README plus configuration, API, patterns and gotchas files where they apply: D1, R2, KV, Queues, Workflows, Hyperdrive, Workers AI, AI Search, AI Gateway, Vectorize, Containers, Browser Rendering, Tunnel, WAF, Pages, Static Assets, Miniflare and more. The D1 reference states the approach: "These references route tasks to the source of truth rather than maintaining copies of APIs, configuration, or plan tables."
It sets working principles. Inspect the project and its pinned versions first, retrieve current docs, "make the smallest change that satisfies the request," and check limits, prices and compatibility flags against current docs because "these can change."
How do you install it?
Cloudflare's README lists several routes. Pick one per agent.
Just this skill, with the skills CLI. This is the command on skills.sh:
npx skills add https://github.com/cloudflare/skills --skill cloudflare
Add -g to install it for all projects. In Claude Code the folder lands in .claude/skills/cloudflare/ in your project, or ~/.claude/skills/cloudflare/ globally, the standard Claude Code skill locations.
Claude Code plugin (every Cloudflare skill plus the MCP server). From the README:
/plugin marketplace add cloudflare/skills
/plugin install cloudflare@cloudflare
This is the better fit for this skill, since it hands off to sibling skills by name. The plugin also bundles Cloudflare's main remote MCP server, which the README describes as giving access to "the Cloudflare API and current developer documentation."
Codex plugin.
codex plugin marketplace add cloudflare/skills
codex plugin add cloudflare@cloudflare
The README says to start a new Codex session afterward.
Cursor, VS Code and manual copy. Cursor installs from its marketplace or as a remote rule pointing at cloudflare/skills. VS Code with GitHub Copilot installs from source after you enable chat.plugins.enabled. You can also clone the repo and copy the folder into your agent's skills directory. For agent-specific details, see Codex skills and Cursor skills.
How do you use it?
Describe the outcome, not the product. The skill is built for requests that never mention Cloudflare product names. Example prompts (suggestions, not quotes from the repo):
I need users to upload profile photos and search them by tag. What should this run on in Cloudflare?Start a new full-stack app on Cloudflare with an API and a small database.We have a Postgres database on another host. How should a Worker talk to it?Add a background job that retries failed invoice emails over a few hours.Build document search over our help center. Should I use AI Search or Vectorize?
Expect a short recommendation first: the products, the role each plays, and the deciding requirement when two could fit. The skill tells the agent to "recommend a small, coherent combination" and to respect your existing stack. Its own example: "a file-upload app can use Workers for its API, R2 for files, D1 for metadata, and Queues for processing."
You need network access for the agent to read the linked docs, and a Cloudflare account and CLI login only when you move from planning to deploying.
Tips and limits
- It is a router, not a manual. Most rows end in a link. Without web access or the Cloudflare MCP server, the agent gets the product map but not the details behind it.
- Install the siblings you will use. It names
wrangler,cf-cli,durable-objects,agents-sdk,turnstile-spinand others. They are optional; without them the agent uses the product docs through the Cloudflare directory. - It checks before promising. The skill tells the agent to "check current availability, limits, and pricing before promising a fit," and flags some products, such as Artifacts, as closed beta.
- No scripts. The folder is Markdown only and runs nothing by itself.
- Audits. On 24 October 2026, skills.sh showed Pass from Gen Agent Trust Hub, Socket and Snyk.
- License. Apache-2.0 lets you copy and change it; keep the license and notices when you redistribute.
Save it in your Kipt library
In Kipt, open Skills and paste the skill's GitHub folder link into the import box. Kipt reads the SKILL.md and the references/ folder; for a repo with many skills, like this one, it shows a list, so you can take wrangler and durable-objects in the same pass. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:
npx skills add https://github.com/<you>/<repo>.git --all -g
Kipt is not affiliated with Cloudflare.
Similar skills
- Wrangler skill: the skill this one hands off to for projects that already use Wrangler.
- Cloudflare Workers skill:
workers-best-practices, for writing and reviewing the Worker code itself. - Durable Objects skill: for the coordination rows, such as chat rooms and bookings.
- Cloudflare Agents SDK skill: for stateful AI agents and remote MCP servers on Workers.
- Turnstile Spin skill, Cloudflare Email Service skill and Cloudflare One skill: the product skills it names for forms, email and Zero Trust.