Wrangler CLI
Guides Wrangler commands and config in projects with wrangler.jsonc or wrangler.toml, checking installed help and current docs and guarding secrets, environments and Previews. Use when deploying, configuring or debugging a Worker with Wrangler.
wrangler- Author
- Cloudflare
- Source
- cloudflare/skills/skills/wrangler @ 5108325
- License
- Apache-2.0
About this skill
What it does
The skill tells the agent how to work with Wrangler in a project that already has wrangler.jsonc, wrangler.json or wrangler.toml. It holds almost no command syntax on purpose: flags and fields change, so the agent reads the project's own wrangler --help, the installed config-schema.json and the current docs page, found through a fifteen-row task table or the Cloudflare MCP docs tool.
Most of SKILL.md is about not breaking things. Before a command the agent pins down the account, Worker, environment and resource, and whether data operations hit local or remote storage. It runs the project's Wrangler without silent upgrades, reconciles dashboard edits a deploy would overwrite, and treats wrangler secret put and secret delete as deployments, since each ships a new version at once. Secrets stay out of arguments and logs. A rollback leaves bound resources and their data as they are.
Workers Previews need project-local Wrangler 4.135.0 or later and, when they target a Wrangler environment, the same --env on every Preview command; production resources get no writes unless you intend them, and the agent warns that Preview URLs are public until access controls are set. Validation is wrangler types and wrangler deploy --dry-run, which proves packaging, not runtime behaviour. references/skill-feedback.md reports a fault in the skill to Cloudflare as a GitHub issue you approve first.
When to use it, and when not
Use it to run, deploy, configure or troubleshoot a Worker in a Wrangler project.
For a new project, one with cloudflare.config.ts or one with no Cloudflare config, the skill steps aside for cf and does not add Wrangler.
How to run it
The install command below puts the folder in place. Wrangler comes from the project's dependencies; remote commands need wrangler login or an API token. To call the skill yourself:
- Claude Code:
/wrangler, or ask for the skill by name. - Codex:
$wrangler. - Cursor: type
/in the Agent chat and pickwrangler.
Example: /wrangler Add a staging environment with its own KV namespace, then deploy a Preview of this branch.
How it differs from similar skills
Cloudflare cf CLI is its counterpart: the config file decides which applies, and each sends the agent to the other CLI. Vercel CLI with tokens spells out concrete commands and asks you for a token when none is in the environment or .env; this skill keeps commands in the docs and never asks for a token in the chat. Deploy to Vercel is a one-step deploy; here a deploy is one task among many.
Install
Run one command — it saves SKILL.md and the files it uses into your agent's skills folder. Claude Code and Codex pick up a new skill without a restart; if it does not show up, restart the agent.
curl -fsSL --create-dirs -o ~/.claude/skills/wrangler/SKILL.md https://anymodel.org/en/skills/cloudflare/wrangler/SKILL.md -o ~/.claude/skills/wrangler/LICENSE https://anymodel.org/en/skills/cloudflare/wrangler/LICENSE -o ~/.claude/skills/wrangler/references/skill-feedback.md https://anymodel.org/en/skills/cloudflare/wrangler/references/skill-feedback.md
curl -fsSL --create-dirs -o ~/.agents/skills/wrangler/SKILL.md https://anymodel.org/en/skills/cloudflare/wrangler/SKILL.md -o ~/.agents/skills/wrangler/LICENSE https://anymodel.org/en/skills/cloudflare/wrangler/LICENSE -o ~/.agents/skills/wrangler/references/skill-feedback.md https://anymodel.org/en/skills/cloudflare/wrangler/references/skill-feedback.md
foreach ($f in 'SKILL.md','LICENSE','references/skill-feedback.md') { $p = Join-Path "$HOME\.claude\skills\wrangler" $f; New-Item -ItemType Directory -Force (Split-Path $p) | Out-Null; Invoke-WebRequest "https://anymodel.org/en/skills/cloudflare/wrangler/$f" -UseBasicParsing -OutFile $p }foreach ($f in 'SKILL.md','LICENSE','references/skill-feedback.md') { $p = Join-Path "$HOME\.agents\skills\wrangler" $f; New-Item -ItemType Directory -Force (Split-Path $p) | Out-Null; Invoke-WebRequest "https://anymodel.org/en/skills/cloudflare/wrangler/$f" -UseBasicParsing -OutFile $p }Any other agent: save the file into its skills folder, or give it the link to the file: https://anymodel.org/en/skills/cloudflare/wrangler/SKILL.md
SKILL.md
Wrangler CLI
Use this skill only when the project has wrangler.jsonc, wrangler.json, or wrangler.toml and no cloudflare.config.ts, or when the user asks for Wrangler. For a new project, a project with cloudflare.config.ts, or a project with no Cloudflare configuration, use the Cloudflare CLI (cf) instead: load the cf-cli skill if it is installed, or follow Use cf with coding agents. Do not add Wrangler to those projects.
Use the project's Wrangler version and retrieve the relevant documentation before writing commands or configuration. CLI flags and configuration fields change; do not rely on memorized examples.
Inspect the Project
- Find the package manager, installed Wrangler version, package scripts, framework, and Wrangler config. Run commands through the project's scripts or package manager so they use its local version. Install dependencies using the existing lockfile when needed; do not silently upgrade Wrangler to match current docs. If the project has a Wrangler configuration file but Wrangler is not a dependency, follow the installation guide to add it locally.
- Identify the config used by the build or deploy command, including framework-generated config. Edit its source rather than generated output.
- Establish the target account, Worker, environment, and resource before running commands that change them. For data operations, determine whether the target is local or remote.
Retrieve What the Task Needs
Use the Cloudflare MCP docs tool if available, or fetch the relevant linked page directly. Follow links to the specific command or product involved; avoid loading the entire reference. If a page moves, rediscover it through the Wrangler command index or Cloudflare docs search.
| Task | Source |
|---|---|
| Discover commands and flags, including resource management, deployments, rollback, and diagnostics | Project-local wrangler --help and wrangler <command> --help; command reference |
| Edit config or add a binding | Installed wrangler/config-schema.json (usually under node_modules); configuration reference |
| Deploy a framework application | Framework guides; follow the guide for the project's existing framework and adapter |
| Migrate an application to Workers when requested | Pages to Workers; Vercel to Workers |
| Configure staging or production | Environments |
| Set secrets locally, in CI, or on a deployed Worker | Secrets |
| Generate binding and runtime types | TypeScript |
| Run locally or choose a testing approach | Local development; testing |
| Choose roles, scopes, or API token permissions for Wrangler | Roles and permissions; Workers roles and permissions |
| Create or manage a branch or pull request environment | Previews overview; get started; configuration |
| Decide whether Preview resources are isolated or shared | Resources and isolation, including its current limitations |
| Configure Preview URLs, access, debugging, or CI | Custom domains; test and debug; examples |
| Choose between Previews, Version URLs, and Wrangler environments | Compare workflows |
| Diagnose authentication or select an account | General commands, including whoami; authentication profiles |
| Deploy an unauthenticated prototype | Claim deployments for eligibility, expiry, and claim URL handling; use a permanent account for production or CI |
Use installed help and schema to check whether documented features exist in the project's version. If a required feature needs an upgrade, make that dependency explicit. If retrieval is unavailable, state the gap and use available local evidence rather than inventing syntax.
Apply the Change
- Prefer
wrangler.jsoncfor new config. Set a new project's compatibility date to today; review runtime changes and test when advancing an existing project's date. Preserve existing project conventions and avoid incidental format migrations. - Check environment inheritance before adding bindings or variables. Some fields must be specified separately for each environment; a working default config does not establish that staging is configured.
- With the Cloudflare Vite plugin, select the environment via
CLOUDFLARE_ENVat dev or build time. Deploy the resulting build; setting an environment at deploy time does not retarget its flattened config. See Vite environments. - Reconcile dashboard changes with the config before deploying: Wrangler can overwrite dashboard variables and routes. When binding existing resources, verify their identifiers; omitted identifiers can trigger automatic provisioning.
- Distinguish local simulation from remote bindings during development. A locally running Worker can still access real resources; check the selected bindings before testing writes.
- Before a remote command, identify the authenticated member or API token and retrieve the current role and scope required for the exact operation. Prefer the narrowest scope that satisfies the user's intent. The
wrangler loginOAuth flow does not support granular authorization; use an account-owned API token when granular access is required, and never ask the user to paste its value into chat. - Keep secret values out of command arguments, source code, and logs. Use the documented interactive input or protected file/stdin mechanism for the command. Local secret files must be ignored by version control and are not automatically uploaded as deployed secrets. For missing local secrets, check file precedence and any
secrets.requireddeclaration in the secrets docs. - Treat
wrangler secret putandsecret deleteas deployments: they create a version and deploy it immediately. Use the documentedwrangler versions secretworkflow when the change must be staged. - Before a rollback, check rollback limitations: connected resources and their data are not rolled back with Worker code.
Work with Workers Previews
- Workers Previews requires project-local Wrangler 4.135.0 or later; project commands do not use a newer global installation. Check the project's pinned version and make any required dependency upgrade explicit.
- Use Previews for branch and pull request environments under the same Worker. Use Version URLs to inspect a specific uploaded version with production resources, and Wrangler environments for persistent, separately deployed Workers. Follow Compare workflows rather than adapting an older aliased Version URL workflow.
- First inspect how the user manages production configuration and deployments. Configure Preview domains through the same configuration system as production, such as Wrangler, the dashboard, or Terraform. Run Preview deployments through the user's existing path, such as the local CLI, Workers Builds, or external CI, unless the user requests a different workflow. Mirror the management approach, not production resource bindings or data.
- Before deploying, follow the current configuration placement table and resource matrix and limitations. Do not infer isolation from the Preview name or allow writes to production resources unless the user explicitly intends to share them.
- Tell the user that Preview URLs are public unless access controls are configured, then let them decide whether to protect the URLs. Follow the custom domain and access guidance for the chosen setup.
- Confirm which Worker or Wrangler environment the user intends to target. When targeting a Wrangler environment, pass the same
--envvalue to every Preview command; omitting it targets the top-level Worker.
Validate
After changing config or bindings in a TypeScript project, regenerate types with the project's wrangler types command rather than hand-editing generated declarations. Run the relevant existing typecheck or tests.
For deployment changes, use the project's build workflow and wrangler deploy --dry-run where supported, with the intended config and environment. A successful dry run checks the build and packaging; it does not prove remote resources or runtime behavior work. Use task-specific local or remote checks as appropriate to the requested work.
For Previews, a returned URL does not validate behavior that depends on bindings Wrangler reports as missing. Validate the behavior the user requested, using Preview-specific logs and configuration when needed. Writes to production resources, destructive testing, and deletion of named Previews must match the user's explicit intent.
Report what changed, the target environment, checks performed, and any unresolved validation gaps. Link the documentation used when the result depends on current command or configuration behavior.
Skill feedback
If the user reports that this skill gave incorrect or outdated guidance, or is missing information, read references/skill-feedback.md.
Connect through AnyModel API
This skill runs on whatever model your agent uses. Point Claude Code, Codex or Cursor at AnyModel to run it on Claude, GPT, Gemini and more with one key — pay per token, no subscription.
AnyModel