agentsclimarketplace

Cloudflare

Skill mbadali25/useful-claude-add-ons/skills/cloudflare

Connect to and work with the Cloudflare v4 API - authenticate (scoped API token or legacy Global API Key), resolve zone/account scope, and read or manage DNS records, zones, cache (purge), firewall/WAF and rulesets, page rules, SSL/TLS settings, DNSSEC, Workers/KV/R2, Zero Trust (Access, Gateway, tunnels), account members, and analytics. Use this skill whenever the user mentions Cloudflare, cloudflare.com, a Cloudflare zone/account, DNS records on Cloudflare, purging Cloudflare cache, WAF/firewall rules, page rules, Workers/KV/R2, Cloudflare Tunnel/Zero Trust/Access/Gateway, or wants to script/report against api.cloudflare.com - even if they don't say "API". Also use it when writing integrations, Terraform-adjacent automation, CI tasks, or one-off queries against the Cloudflare API, and when debugging 400/403/429/9xxx errors from api.cloudflare.com.From its SKILL.md

Install
npx -y skills add mbadali25/useful-claude-add-ons --skill cloudflare

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

8.6 KB, as published. Nobody here has run it

Cloudflare v4 API

Automate and query Cloudflare: DNS, zones, cache purge, firewall/WAF and rulesets, page rules, SSL/TLS, DNSSEC, Workers/KV/R2, Zero Trust (Access, Gateway, tunnels), members, and analytics. Everything is RESTful JSON over HTTPS against a single host: https://api.cloudflare.com/client/v4.

There is no regional host and no "app" host to confuse - every call goes to api.cloudflare.com/client/v4. The dashboard at dash.cloudflare.com is not an API endpoint.

Quick start

  1. Get credentials (see references/auth.md). Strongly prefer a scoped API token (My Profile -> API Tokens -> Create Token) over the legacy Global API Key. A token carries only the permissions and zone/account scope you grant it; the Global API Key is root-equivalent over the whole account and can't be scoped.

  2. Verify the token before doing anything else - this is the fastest way to catch a bad key or wrong scope:

    export CLOUDFLARE_API_TOKEN=xxxxx
    python scripts/cloudflare_client.py verify        # GET /user/tokens/verify
    
  3. Figure out scope. Almost every resource hangs off one of two prefixes:

    • Zone-scoped: /zones/{zone_id}/... (DNS records, cache, page rules, most WAF, SSL settings, DNSSEC).
    • Account-scoped: /accounts/{account_id}/... (Workers, R2, KV, Zero Trust, account rulesets, members).

    Set the ids in the environment and skip the lookup:

    export CLOUDFLARE_ACCOUNT_ID=...   # dashboard -> the account -> id in the URL / sidebar
    export CLOUDFLARE_ZONE_ID=...      # dashboard -> the zone -> Overview -> API section
    

    This is not a convenience. A scoped token frequently cannot list /accounts or /zones at all - it returns 403, or an empty array, even when it has full permission on the account you care about. So resolving an id by name fails in a way that looks like a broken token and is not. Set the ids, and nothing has to be enumerated.

    Resolve by name only when the credential is allowed to enumerate: python scripts/cloudflare_client.py zone-id example.com. With one visible zone or account, the name can be omitted entirely.

Helper script

scripts/cloudflare_client.py is a self-contained Python client (stdlib only - no third-party packages) that handles both auth methods, the standard {success, errors, messages, result, result_info} envelope, offset (page/per_page) pagination, 429 backoff (Cloudflare does not send Retry-After, so it uses exponential backoff), and a read-only / dry-run guard for mutations. Use it instead of re-writing boilerplate:

export CLOUDFLARE_API_TOKEN=xxxxx                 # scoped token - use this
export CLOUDFLARE_ACCOUNT_ID=xxxxx                # optional, skips a lookup a token may not be allowed
export CLOUDFLARE_ZONE_ID=xxxxx                   # optional, same
# export [email protected] CLOUDFLARE_API_KEY=global_key   # legacy, root-equivalent, avoid

python scripts/cloudflare_client.py verify                                   # token check
python scripts/cloudflare_client.py zone-id example.com                      # name -> zone_id
python scripts/cloudflare_client.py get     /zones/{zid}/dns_records --params 'type=A' 'name=www.example.com'
python scripts/cloudflare_client.py get-all /zones/{zid}/dns_records         # auto-paginate every page
python scripts/cloudflare_client.py post    /zones/{zid}/dns_records --json '{"type":"A","name":"www","content":"1.2.3.4","ttl":300,"proxied":true}'
python scripts/cloudflare_client.py put      /zones/{zid}/dns_records/{rid} --json '{"type":"A","name":"www","content":"5.6.7.8","proxied":true}' --dry-run
python scripts/cloudflare_client.py delete   /zones/{zid}/dns_records/{rid}

Set CLOUDFLARE_READ_ONLY=1 to block every mutating call, or pass --dry-run on a single mutation to print the exact request without sending it. Importable too: from cloudflare_client import CloudflareClient.

Response envelope

Every response uses the same wrapper - check success before trusting result:

{ "success": true, "errors": [], "messages": [],
  "result": { ... } or [ ... ],
  "result_info": { "page": 1, "per_page": 100, "count": 100, "total_count": 342, "total_pages": 4 } }

On failure success is false and errors holds [{ "code": 81044, "message": "..." }]. Cloudflare error codes are numeric and product-specific - surface the code and message verbatim to the user; don't guess. result_info is present only on list endpoints, and that's what get-all keys off to know when to stop.

API map - which reference to read

TaskReference file
Token vs Global Key, creating/scoping tokens, verify, permission groups, secret handlingreferences/auth.md
Zone vs account scoping, resolving IDs, key resource paths (DNS, cache purge, firewall/rulesets, page rules, SSL/TLS, DNSSEC, Workers/KV/R2, Zero Trust, members), pagination limits per resourcereferences/api-map.md
Recipes: bulk DNS export/import, targeted vs full cache purge, DNS upsert, mint a least-privilege token, list all zones in an account, find a proxied record's real originreferences/common-tasks.md

Read the relevant reference before writing calls - they carry the exact paths, per-page caps, and object shapes.

Universal conventions

  • Auth header: scoped token -> Authorization: Bearer <token>. Legacy -> X-Auth-Email: <email> + X-Auth-Key: <global_key>. Never mix the two on one request. The helper picks based on which env vars are set (token wins if both present).
  • Rate limit: 1,200 requests per 5-minute rolling window per token/user, across all endpoints. There is no Retry-After header on 429 - back off exponentially (the helper does this). Throttle bulk exports and batch DNS work.
  • Pagination: offset-based ?page=1&per_page=N. The per-page cap varies by resource (DNS records allow up to 5,000 with per_page; many endpoints cap at 50 or 100). Read result_info.total_pages and loop - get-all handles this. A few newer endpoints (some Zero Trust, Workers) use cursor; the helper passes a cursor param straight through if you supply one.
  • IDs are opaque 32-char hex strings. A "zone id" is not the domain name; a member id is the membership id, not the user id. Always resolve names to IDs first.
  • Timestamps: ISO 8601 UTC.
  • Proxied vs DNS-only: for DNS records, "proxied": true routes traffic through Cloudflare (orange cloud) and hides the origin IP; false is DNS-only (grey cloud). Changing this flag changes production traffic behaviour - treat it as a mutation that matters.

Safety rails for writes

Cloudflare changes take effect on live production traffic within seconds - a bad DNS edit, an over-broad cache purge, or a firewall rule change can cause an outage. Before any POST/PUT/PATCH/DELETE:

  • Confirm the exact zone/account and the exact record with the user; show the record(s) the filter matched, not just a count, when feasible.
  • Default to dry-run: use --dry-run (or CLOUDFLARE_READ_ONLY=1) to preview, then act only on explicit confirmation.
  • Be especially careful with: DNS record edits/deletes (can take a site offline or misroute mail - check MX/TXT/SPF before touching), proxied flips (exposes origin IP or breaks WAF coverage), full cache purge (purge_everything - prefer purging specific URLs/tags/hosts to avoid an origin load spike), firewall/WAF rule and ruleset changes (can lock out real users or open a hole), and DNSSEC toggles (can break resolution if the registrar DS record isn't kept in sync).
  • Prefer least-privilege when helping create a token: grant only the specific permission groups and the specific zones/account the task needs (see references/auth.md), never the Global API Key for automation.

Keep looking

Skills are one crate of 609,856. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.