cloudflare/workers-sdk wrangler@4.97.0

on GitHub

Minor Changes

• #13996 94b29f7 Thanks @vaishnav-mk! - Add restart-from-step options to wrangler workflows instances restart

  You can now restart a Workflow instance from a specific step using --from-step-name, with optional --from-step-count and --from-step-type disambiguation. These options work for both remote Workflow instances and local wrangler dev --local sessions.

Patch Changes

• #14141 b210c5e Thanks @MattieTK! - Add re-authentication hint to account fetch error messages

  When Wrangler fails to automatically retrieve account IDs, the error messages now suggest running wrangler login as a troubleshooting step. This addresses confusion for users who encounter these errors after OAuth system changes or other authentication issues.

• #14078 aec1bb8 Thanks @MattieTK! - Bump am-i-vibing from 0.1.1 to 0.4.0

  This updates the agentic environment detection library to the latest version, which includes improved detection coverage for newer AI coding agents.

• #14147 e06cbb7 Thanks @dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260529.1	1.20260601.1

• #14027 9a26191 Thanks @matingathani! - Gracefully handle EMFILE error when assets directory exceeds OS watcher limit

  Previously, when wrangler dev was pointed at an assets directory with more than ~4,096 subdirectories, the chokidar file watcher threw an EMFILE: too many open files error that was not caught, causing an infinite error loop that made the dev server unresponsive.

  Now the error is caught and wrangler:

  1. Logs a clear warning explaining the platform watcher limit was hit
  2. Recommends reducing the number of subdirectories by flattening or restructuring the assets directory
  3. Disables the assets watcher gracefully so the dev server continues working without hot-reload

• #14041 5565823 Thanks @matingathani! - Fix wrangler complete printing the AI skills prompt into shell completion output

  Previously, running eval "$(wrangler complete zsh)" (or any other shell) would fail with errors like zsh: command not found: --install-skills because the interactive AI agent skills installation prompt was included in the completion script output.

  The skills prompt is now skipped when running wrangler complete, so the generated completion script is clean and can be sourced correctly.

• #13881 890fca7 Thanks @matingathani! - Show a clear error when --metadata is not valid JSON instead of silently ignoring the value

• #14149 6fc9777 Thanks @mattjohnsonpint! - Fix wrangler deploy --upload-source-maps silently skipping source maps when the entry file ends with magic comments after //# sourceMappingURL=

  Wrangler previously assumed the //# sourceMappingURL= comment was the last non-empty line of a module. Tools like sentry-cli sourcemaps inject append a //# debugId= comment after it, which silently caused source maps to be omitted from the upload form, most commonly when deploying with --no-bundle --upload-source-maps. Wrangler now scans trailing magic comments (lines starting with //# or //@) and detects the //# sourceMappingURL= comment regardless of which other magic comments follow it.

• #14105 337e912 Thanks @dario-piotrowicz! - Remove trailing periods from URLs in terminal output

  URLs printed to the terminal with a sentence-ending period (e.g. https://example.com/path.) would include the period when clicked in some terminal emulators, causing 404 errors. This removes trailing periods from all URLs displayed in CLI output across wrangler, miniflare, vitest-pool-workers, and workers-utils.

• #14150 8e7b74f Thanks @avenceslau! - Fix Workflows schedules deploy payload to match the control plane API

  When deploying a Workflow with a schedules binding property, Wrangler sent the cron expressions as a list of strings. The Workflows API expects a list of objects of the form { cron: string }, so the request was rejected. Wrangler now maps each configured cron expression to { cron } (normalizing a single string or an array) when building the request. The user-facing config still accepts a string or an array of strings.

• #14084 e86489a Thanks @dario-piotrowicz! - Fix JSON variable bindings in wrangler init --from-dash and remote config diff

  When fetching a remote Worker's configuration, JSON variable bindings (e.g. {"my_value": 5}) were incorrectly serialized as { "name": "MY_JSON", "json": {"my_value": 5} } instead of { "MY_JSON": {"my_value": 5} }. This affected two areas:

  • wrangler init --from-dash would generate a wrangler.json with broken vars entries
  • Remote config diff checks would always report JSON bindings as changed, since the malformed remote representation could never match the local config

  Both issues are now fixed and remote JSON bindings are now correctly mapped.

• #14155 42288d4 Thanks @dario-piotrowicz! - Include agent skill installation status in all telemetry events

  The agent skill installation status is now consistently included in all telemetry events, not just a subset of them.

• #14063 65b5f9e Thanks @emily-shen! - Move fetch helpers into @cloudflare/workers-utils

  Shared Cloudflare API fetch helper types and plumbing now live in @cloudflare/workers-utils so Wrangler and other clients can use the same implementation.

• #14112 3a746ac Thanks @penalosa! - Pin non-bundled runtime dependencies to exact versions

  Dependencies that are not bundled into a package's published output are installed directly into consumers' dependency trees, so they are now pinned to exact versions instead of semver ranges. This closes a supply-chain gap where an unpinned external dependency could resolve to a compromised upstream release on a fresh install. A new pnpm check:pinned-deps lint enforces this for all published packages (and for the shared pnpm catalog) going forward.

• #14124 64ef9fd Thanks @odiak! - Fix wrangler secret bulk dropping newlines from .env input read from stdin

  Previously, .env input piped through stdin was concatenated without line breaks, so only the first secret could be parsed correctly. Stdin input now preserves line separators before parsing.

• Updated dependencies [e06cbb7, 4ef790b, 337e912, 3a746ac]:

  • miniflare@4.20260601.0