Skip to content
Skills Library
DevOps & cloud · Tool integration

cloudflare

/cloudflare

Manage Cloudflare from the command line without an MCP server, using the two installed binaries: `cf` for the account API surface (zones, DNS, firewall, rulesets, KV, R2, D1, Queues, Pages, Workers metadata, Zero Trust) and `wrangler` for the Workers project loop (dev, deploy, tail, secrets). Use when adding or editing DNS records, deploying or tailing a Worker, inspecting a binding's stored data, or auditing what a Cloudflare account currently holds.

Licence
MIT
Last repository change
10.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
4
Deprecation

No signals detected. Repository changed on 10.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Manage Cloudflare from the command line without an MCP server, using the two installed binaries: `cf` for the account API surface (zones, DNS, firewall, rulesets, KV, R2, D1, Queues, Pages, Workers metadata, Zero Trust) and `wrangler` for the Workers project loop (dev, deploy, tail, secrets). Use when adding or editing DNS records, deploying or tailing a Worker, inspecting a binding's stored data, or auditing what a Cloudflare account currently holds.

Original author description

Manage Cloudflare from the command line without an MCP server, using the two installed binaries: `cf` for the account API surface (zones, DNS, firewall, rulesets, KV, R2, D1, Queues, Pages, Workers metadata, Zero Trust) and `wrangler` for the Workers project loop (dev, deploy, tail, secrets). Use when adding or editing DNS records, deploying or tailing a Worker, inspecting a binding's stored data, or auditing what a Cloudflare account currently holds.

How it works

  1. Auth
  2. Before any write
  3. cf: the account
  4. wrangler: the Workers project
  5. Not installed
  6. Reference
Original SKILL.md headings · 119 lines, 1,347 words
SKILL.md preview
---
name: cloudflare
description: >-
  Manage Cloudflare from the command line without an MCP server, using the two installed binaries:
  `cf` for the account API surface (zones, DNS, firewall, rulesets, KV, R2, D1, Queues, Pages,
  Workers metadata, Zero Trust) and `wrangler` for the Workers project loop (dev, deploy, tail,
  secrets). Use when adding or editing DNS records, deploying or tailing a Worker, inspecting a
  binding's stored data, or auditing what a Cloudflare account currently holds.
model: sonnet
effort: medium
# Reads only. Every mutating command is deliberately absent so it still hits the
# permission prompt, which is the same confirmation the "Before any write" section
# below asks for.
allowed-tools:
  - Bash(cf auth list*)
  - Bash(cf auth whoami*)
  - Bash(cf user tokens verify*)
  - Bash(cf zones list*)
  - Bash(cf zones get*)
  - Bash(cf dns records list*)
  - Bash(cf dns records get*)
  - Bash(cf firewall access-rules list*)
  - Bash(cf rulesets list*)
  - Bash(cf kv namespaces list*)
  - Bash(cf r2 buckets list*)
  - Bash(cf d1 list*)
  - Bash(cf queues list*)
  - Bash(cf pages projects list*)
  - Bash(cf workers scripts list*)
  - Bash(cf zero-trust tunnels list*)
  - Bash(cf schema*)
  - Bash(cf agent-context*)
  - Bash(wrangler whoami*)
  - Bash(wrangler deployments list*)
  - Bash(wrangler secret list*)
  - Bash(jq *)
  - Read
---

# Cloudflare from the CLI

Two binaries, and the split is about *where the state lives* rather than which product it belongs to.

| You are touching | Tool |
|---|---|
| Anything that is a row in the Cloudflare account: zones, DNS, firewall, rulesets, KV namespaces, R2 buckets, D1 databases, Queues, Pages projects, Zero Trust, tokens | `cf` |
| Anything that reads the project on disk: local dev server, deploying the Worker in cwd, live logs, a Worker's secrets and bindings | `wrangler` |

The overlap is real and not a problem: `cf d1 list` and `wrangler d1 list` both work. Reach for `cf` when you want the account's answer as JSON, and `wrangler` when the command only makes sense inside a Workers project. The formula is `cloudflare-wrangler` but the binary is `wrangler`; `cf` is a bun global install (`bun add -g cf`). Both are wired up by the `apps` role.

`cf` is a Cloudflare-published technical preview that is being built to *become* wrangler, so expect its commands to move. Check `cf --help` before trusting a command shape from memory, including the ones below.

## Auth

Nothing is exported into the shell.
This machine works across several Cloudflare accounts, so each tool holds its own credentials and the account in play is a property of where you are, not of the environment.

`cf` resolves credentials in one order: `CLOUDFLARE_API_TOKEN` first, then the profile named by `--profile`, then the profile bound to the nearest parent directory, then `default`.
That order is why the env var is absent rather than just unused: while it is set, `cf auth login` refuses to run and every `--profile` is ignored, so one token would quietly outrank the account asked for.

```bash
cf auth list                                     # profiles, and which is bound here
cf auth whoami                                   # account behind the current credentials
cf auth create <name>                            # browser OAuth, stored under that name
cf auth activate <name> ~/Developer/<project>    # bind a profile to a directory
cf zones list --profile <name>                   # one-off override
```

Prefer a directory binding over `--profile` for project work, and `--profile` when one session crosses accounts.
When a read comes back with plausible data for the wrong account, `cf auth whoami` is the first thing to check, ahead of the command itself.

`wrangler` has no profile system.
It has the single OAuth session from `wrangler login`, and it takes the account from `account_id` in the project's `wrangler.jsonc`.
Set that field in any project outside the default account rather than exporting `CLOUDFLARE_ACCOUNT_ID`, and switch accounts with `wrangler logout` then `wrangler login`.
Both tools store this under `~/Library/Preferences/.wrangler`, which the agent sandbox denies reading, so credentials stay out of anything an agent can see. Under Claude Code `wrangler` and `cf` are exempt from the sandbox as the leading token of a call, so they read those credentials natively; that exemption carries no allow rule, so every call still asks. Run each as its own call: piped or chained, it is sandboxed and fails to authenticate.

Scope failures read differently under OAuth.
A `403` on one subcommand while others work still means missing permission rather than broken auth, but with a profile it is the account's role, not a token scope, so say which one you think it is instead of retrying the same call.
`cf user tokens verify` only answers for an API token, so it applies to a CI token or a token handed over for one task, not to a profile.

**Never suggest the Global API Key.** It authenticates the entire account including billing and cannot be scoped. If a task genuinely needs permission the current credentials lack, ask the user for a profile on an account that has it or a scoped token minted for it; do not route around the restriction.

## Before any write

Cloudflare writes hit live DNS, so the blast radius of a mistake is "the domain is down" and the feedback loop is a TTL. `cf` gives you two safety rails the older CLIs never had, so use both:

1. **`--dry-run` first.** It validates the call and shows what would happen without executing. Run it on every mutating command and show the user the output.
2. **Never pass `-f` / `--force`.** Destructive `cf` commands prompt for confirmation on their own, and `--force` exists to skip that in CI. Letting the prompt reach the user is the point.
3. **List before you write.** Record IDs are opaque, are not stable across a delete and recreate, and are the only handle `dns records edit` and `dns records delete` accept. Never carry an ID over from an earlier session.
4. **Never guess a zone.** `-z` takes a zone ID *or* a domain name, so there is no reason to resolve one by hand, but if more than one zone plausibly matches what the user said, ask.
5. **State the proxy mode you are setting.** Proxied (orange cloud) versus DNS-only (grey cloud) decides whether traffic passes through Cloudflare, which changes the visible IP and the TLS terminator. Flipping it on an existing record is a user-visible change even though the record content is untouched.

## cf: the account

Every command takes `-z <zone-id-or-domain>` where a zone is relevant, and **all output is JSON on stdout** with status messages on stderr, so piping into `jq` always works.

```bash
cf zones list | jq -r '.[] | "\(.name)\t\(.id)"'
cf zones get -z example.com
cf dns records list -z example.com | jq -r '.[] | "\(.type)\t\(.name)\t\(.content)\t\(.id)"'
cf dns records list -z example.com --name-contains api
```

```bash
# writes: dry-run first, then the same command without it, and let cf prompt
cf dns records create -z example.com --dry-run --type A --name www --content 203.0.113.10
cf dns records edit -z example.com <dns-record-id> --content 203.0.113.11
cf dns records delete -z example.com <dns-record-id>
```

```bash
cf firewall access-rules list -z example.com
cf rulesets list -z example.com
cf kv namespaces list
cf r2 buckets list
cf d1 list
cf queues list
cf pages projects list
cf workers scripts list
cf zero-trust tunnels list
```

List endpoints paginate. A read that silently stops at the first page and gets reported as the full picture is the most common way an audit here goes wrong, so pass `--per-page` and keep going while a page comes back full.

Two commands worth knowing when a command shape is unclear: `cf schema <command>` prints the API schema behind it, and `cf agent-context [product]` prints Cloudflare's own context for a product. Prefer either over guessing flags.

## wrangler: the Workers project

```bash
wrangler whoami
wrangler dev                        # local dev server; press `e` for the Local Explorer
wrangler deploy                     # deploys the project in cwd
wrangler tail <worker-name>         # live request logs
wrangler deployments list
wrangler rollback                   # previous deployment, confirm first
wrangler secret list --name <worker>
wrangler pages deploy ./dist --project-name <project>
wrangler d1 execute <db> --command "select 1"        # add --remote to hit production
```

Two traps worth naming. `wrangler d1 execute` runs against the **local** simulated database unless you pass `--remote`, so a query that "returns nothing" is usually pointed at the wrong place. And `wrangler deploy` reads `wrangler.jsonc` (or `wrangler.toml`) from the current directory, so it is only meaningful inside a Workers project; run it anywhere else and it fails on missing config rather than doing nothing.

The Local Explorer in `wrangler dev` is the fastest way to see what a binding actually holds locally, so reach for it before writing a script to introspect `.wrangler/state`.

## Not installed

Deliberately absent, so do not reach for them or suggest they should already be there:

- **`cloudflared`**, the tunnel daemon. `cf zero-trust tunnels list` manages tunnel *records*, but actually carrying traffic needs the daemon. If the user wants a tunnel running, that is `brew install cloudflared` plus a one-time `cloudflared tunnel login`, and it is their call to make.
- **`cf-terraforming`** and OpenTofu, for exporting live config into IaC. Say so if a task wants Cloudflare managed declaratively; do not improvise an export.

## Reference

- [cf on npm](https://www.npmjs.com/package/cf) and the [announcement](https://blog.cloudflare.com/cf-cli-local-explorer/)
- [wrangler commands](https://developers.cloudflare.com/workers/wrangler/commands/)
- [API v4](https://developers.cloudflare.com/api/)

Versions

RepositoryScoreLicence
This page · jmanuelrosa/agent-catalog100MIT
RIJUDEVx/agent-skills100MIT
ericrisco/rsc-harness100MIT
travnie/twojstar100ISC
getsentry/junior100Apache-2.0
tdimino/claude-code-minoan100MIT
lynricsy/hyperskills100MIT
Audaa-Cyber/orchill100Apache-2.0
cloudflare/skills100Apache-2.0

Database snapshot: 13.11.2026 · package version 5ab4c38306bb (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100cloudflare-deployment
    Deploy and manage Cloudflare Workers, Pages, KV, R2, and D1 using wrangler CLI or MCP server. Use when working with Cloudflare services, serverless functions, or edge deployments.
  • 100wrangler
    Deploy and manage Cloudflare Workers, Pages, R2, D1, KV. Use when working with wrangler, Cloudflare Workers, or edge deployments.
  • 100cloudflare-ops
    Cloudflare Workers + Wrangler edge ops: runtime, bindings, local dev, secrets, deploy/CI, Pages-vs-Workers. Triggers on: cloudflare workers, wrangler, wrangler deploy, wrangler.toml, KV, D1, R2, durable objects, queues, vectorize, compatibility_date, edge functions, illegal…
  • 100cloudflare-cf-cli
    Cloudflare's official cf CLI (beta) for the public Cloudflare API and Workers projects. Use for finding and running Cloudflare commands from the terminal — sign-in and profiles, cf cli search, cf schema, cf deploy/dev/build, resource commands for DNS, zones, D1, KV, and R2 — and…
  • 95cloudflare-manager
    Comprehensive Cloudflare account management for deploying Workers, KV Storage, R2, Pages, DNS, and Routes. Use when deploying cloudflare services, managing worker containers, configuring KV/R2 storage, or setting up DNS/routing. Requires CLOUDFLARE_API_KEY in .env and Bun…
  • 100cloudflare-manager
    Comprehensive Cloudflare account management for deploying Workers, KV Storage, R2, Pages, DNS, and Routes. Use when deploying cloudflare services, managing worker containers, configuring KV/R2 storage, or setting up DNS/routing. Requires CLOUDFLARE_API_KEY in .env and Bun…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.