Update all non-major dependencies

[renovate]

ℹ️ Note

This PR body was truncated due to platform limits.

This PR contains the following updates:

Package	Change	Age	Confidence
@opennextjs/cloudflare (source)	^1.19.10 → ^1.19.11
@types/node (source)	^25.8.0 → ^25.9.1
@types/react (source)	^19.2.14 → ^19.2.15
eslint (source)	^9 → ^9.39.4
pnpm (source)	11.1.2 → 11.3.0
postcss (source)	^8.5.14 → ^8.5.15
wrangler (source)	^4.91.0 → ^4.94.0

---

Release Notes

opennextjs/opennextjs-cloudflare (@​opennextjs/cloudflare)

v1.19.11

Compare Source

Patch Changes

• #​1270 802047e Thanks @​alex-all3dp! - fix: skip non-upload-triggered worker versions when building skew-protection deployment mapping

  Worker versions created by metadata-only operations (e.g. Cloudflare API secret updates) do not include the static assets bundle. Previously, such versions could become the "latest" target in the skew-protection mapping, causing /_next/static/* requests to return 404 on past deployments. Versions are now filtered to those with workers/triggered_by in {upload, version_upload}.

  Closes #​1230

pnpm/pnpm (pnpm)

v11.3.0

Compare Source

Minor Changes

• Added pnpm stage with publish, list, view, approve, reject, and download subcommands for npm staged publishing.

• Added a new setting trustLockfile. When true, pnpm install skips the supply-chain verification pass that re-applies minimumReleaseAge / trustPolicy='no-downgrade' to every entry in the loaded lockfile. The install treats the lockfile as already-trusted — useful for closed-source projects where every commit comes from a trusted author. Defaults to false; verification stays on by default. Set in pnpm-workspace.yaml.

  Also cut the memory footprint of the verification pass itself: the per-(registry, name) trust-meta cache previously retained the full packument — dependency graphs, scripts, README, and per-version manifests — for the entire install. On large workspaces (~4k lockfile entries with minimumReleaseAge + trustPolicy: no-downgrade enabled) this could OOM CI runners with a 2GB heap cap. The cache now stores only the fields the trust check actually reads (time, per-version _npmUser.trustedPublisher, dist.attestations.provenance). The abbreviated-metadata cache is similarly projected to just the package-level modified field and the set of currently-listed version names. Fixes #​11860.

• Implemented pnpm pkg command natively, following npm pkg standards.

• Implemented pnpm repo command natively, following npm repo standards.

• Implemented pnpm set-script (alias ss) natively. Adds or updates an entry in the scripts field of the project manifest, supporting package.json, package.json5, and package.yaml formats.

• Add a skip-manifest-obfuscation option for pnpm pack and pnpm publish. When enabled, the original packageManager field and publish lifecycle scripts are kept in the packed/published manifest instead of being stripped. The pnpm-specific pnpm field continues to be omitted.

Patch Changes

• Fixed pnpm dlx failing with ERR_PNPM_NO_IMPORTER_MANIFEST_FOUND when the installed package's CAS slot is missing its package.json. Observed in the wild for pnpm dlx node@runtime:<version> when the GVS slot was populated without the synthesized manifest runtime archives need (they don't ship a package.json of their own, so the synthesized one is the only way it gets there; an existing slot from an earlier code path that skipped the synthesis stays incomplete). The bin link itself is wired up from the resolution and remains valid, so dlx now falls back to the scopeless package name when the slot's manifest is unreadable — for single-bin packages (the dlx common case, including every runtime: spec) this matches what manifest.bin would have named. Multi-bin packages already require --package=<spec> <bin> to disambiguate and don't enter this code path.
• Fixed non-determinism in pnpm dedupe and pnpm install when a dependency graph contains packages with transitive peer dependencies on each other (e.g. @aws-sdk/client-sts and @aws-sdk/client-sso-oidc) and auto-install-peers is enabled. The lockfile no longer flips between two equally-valid forms across consecutive runs. The root cause was that resolveDependencies pushed onto its pkgAddresses / postponedResolutionsQueue arrays from inside Promise.all-spawned callbacks, so completion-order timing leaked into the array order and downstream cyclic-peer suffix assignment. Fixes #​8155.
• Fixed a regression introduced by #​11711 where pnpm add <github-shorthand> (and any other wanted-dependency whose alias can't be parsed from the user-supplied spec, e.g. tarball URLs or pnpm/test-git-fetch#sha) was silently dropped from the manifest update and from pendingBuilds. The alias-keyed lookup added in that PR couldn't find a wantedDependency whose alias was undefined at parse time but resolved to a package name only after fetching, so the entry never made it into specsToUpsert. Restored the original index-based pairing between directDependencies and wantedDependencies; the catalog-protocol preservation that PR was originally fixing is unaffected because it's driven by rdd.catalogLookup.userSpecifiedBareSpecifier, not by the lookup. Fixes the three rebuilds dependencies / rebuilds specific dependencies / rebuild with pending option failures in building/commands/test/build/index.ts.
• Fixed pnpm add --config leaving orphan entries in pnpm-lock.env.yaml (the optional subdependencies of the previously resolved version of the updated config dependency).

v11.2.2

Compare Source

Patch Changes

• When the install engine is delegated to pacquet via configDependencies, the user's CLI flags passed to pnpm install (e.g. --no-runtime, --prod, --dev, --no-optional, --node-linker, --cpu/--os/--libc, --offline, --prefer-offline) are now forwarded to pacquet's install subcommand verbatim. Previously pacquet was invoked with a fixed argument list, so flags like --no-runtime were silently dropped. Flag forwarding is gated on the command being install/i; add, update, and dedupe still don't forward (their flag surface doesn't line up with pacquet's install).
• Fixed pnpm up (and pnpm add / pnpm remove) failing with pacquet_package_manager::outdated_lockfile when pacquet is declared in configDependencies. pnpm now passes --ignore-manifest-check to pacquet so its --frozen-lockfile check doesn't fire against the (pre-mutation) package.json pnpm hasn't written yet #​11797. Requires a pacquet release that supports the flag — bump PACQUET_VERSION in the e2e tests once it ships.

v11.2.1

Compare Source

Patch Changes

• Mark optional subdependency snapshots of config dependencies with optional: true in the env lockfile, matching how optional dependencies are recorded elsewhere in pnpm-lock.yaml. Previously, snapshots for the platform-specific subdeps pulled in via a config dep's optionalDependencies were written as empty objects, which was inconsistent with the rest of the lockfile and made it look like those non-host platform variants were required.
• Fix pickRegistryForPackage returning the wrong registry for an unscoped npm: alias under a scoped local name. A manifest entry like "@​private/foo": "npm:lodash@^1" was routing the lodash fetch through registries["@​private"], even though lodash is unscoped and doesn't live on that registry. The npm-alias branch now returns the alias target's own scope (or null for an unscoped target, falling through to registries.default) instead of leaking into the local key's scope.
• Don't print "Installing config dependencies..." when config dependencies are already installed and nothing needs to be fetched, re-linked, or removed.

v11.2.0

Compare Source

Minor Changes

• Experimental: Adding @pnpm/pacquet (the Rust port of pnpm) to configDependencies in pnpm-workspace.yaml now delegates the materialization phase of pnpm install to the pacquet binary. pnpm still owns dependency resolution; pacquet only fetches and imports from the freshly-written lockfile. This is an opt-in preview of the Rust install engine #​11723.

  To configure pacquet in a project, run:

  pnpm add @​pnpm/pacquet --config
  

  You'll see changes in pnpm-workspace.yaml and pnpm-lock.yaml that should be committed. If you experience any issues with pacquet, please let us know by mentioning this in the GitHub issue you create.

• configDependencies now resolve and install one level of optionalDependencies declared by the config dependency, with os/cpu/libc platform filtering applied at install time. This unlocks the esbuild/swc-style pattern where a package ships platform-specific binaries via optionalDependencies — a config dependency can now do the same and have the matching binary symlinked next to it in the global virtual store, so require('pkg-platform-arch') from inside the config dependency resolves correctly.

  The env lockfile records all platform variants regardless of host platform, so it remains portable across machines. Each entry in a config dependency's optionalDependencies must declare an exact version — ranges and tags are rejected to keep installs reproducible.

• Implement the documented pnpm login --scope <scope> flag. The scope is normalized (a leading @ is added if missing; blank values are ignored) and an @<scope>:registry=<registry> mapping is written to the pnpm auth file alongside the auth token. Subsequent installs of @<scope>/* packages then route to the chosen registry. Previously pnpm login --scope foo errored with Unknown option: 'scope' despite the flag being listed in the online documentation #​11716.

• pnpm outdated and pnpm update --interactive now report Node.js, Deno, and Bun runtimes installed as project dependencies (runtime: specifiers). Previously these were silently skipped.

Patch Changes

• Fix cafile=<relative-path> in .npmrc being read from the wrong directory when pnpm is invoked from a different cwd (e.g. pnpm --dir <project> install from a CI wrapper or monorepo script). The path is now resolved against the directory of the .npmrc that declared it, not process.cwd(). Before this fix the CA file silently failed to load — the install proceeded without the configured CA and the user only saw TLS errors against a private registry, with no log line tying back to the wrongly resolved path #​11624.

• Fix config.registry getting a trailing slash appended when registry is set in .npmrc and no registries.default is provided by pnpm-workspace.yaml. The sync from registries.default to config.registry introduced in #​11744 now only fires when the workspace manifest actually contributes a different default.

• Fix global add/update to handle minimumReleaseAge policy violations instead of surfacing an internal resolver guardrail error.

• Fix two crashes with injectWorkspacePackages: true when the lockfile has been pruned (e.g. by turbo prune --docker):

  • Cannot use 'in' operator to search for 'directory' in undefined: a peer-dependency-variant injected snapshot inherits its resolution from the base packages: entry; when a pruner drops that base entry the readers crash. convertToLockfileObject now reconstructs the directory resolution from the file: depPath at load time — a single normalization point, so every reader sees a fully-formed snapshot.
  • ERR_PNPM_ENOENT on node_modules/.bin/<tool>: after prepare/postinstall, runLifecycleHooksConcurrently re-imported each injected workspace package; the scanDir-into-filesMap workaround fed target-internal paths to the importer, which the makeEmptyDir fast path (#​11088) then wiped. Drop the workaround and pass keepModulesDir: true so the importer preserves the target's existing node_modules (bin links + transitive deps) and source files keep their hardlinks.

• Fixed pnpm login and pnpm logout ignoring registries.default from pnpm-workspace.yaml #​10099.

• Fix the minimumReleaseAge (publishedBy) maturity shortcut to be inclusive at the cutoff. Previously, abbreviated metadata whose modified field equalled the cutoff fell off the fast path and triggered a full-metadata re-fetch (or a MISSING_TIME error when full metadata wasn't permitted). Since modified is an upper bound on every version's publish time, modified == publishedBy already implies every version passes the per-version <= filter in filterPkgMetadataByPublishDate, so the shortcut now accepts the boundary case directly. Strictly > (was >=) at the rejection branch.

• Honor publishConfig.access when publishing packages.

v11.1.3

Compare Source

Patch Changes

• pnpm install now re-validates pnpm-lock.yaml entries against the active minimumReleaseAge and trustPolicy: 'no-downgrade' policies before any tarball is fetched. Lockfiles resolved elsewhere (committed to the repo, restored from a CI cache, produced by an older pnpm) under a weaker or absent policy can no longer install a freshly-published or trust-downgraded version silently. Violating entries abort the install with ERR_PNPM_MINIMUM_RELEASE_AGE_VIOLATION, ERR_PNPM_TRUST_DOWNGRADE, or the generic ERR_PNPM_LOCKFILE_RESOLUTION_VERIFICATION when both policies trip in the same batch; minimumReleaseAgeExclude and trustPolicyExclude are honored. Verification results are cached so repeat installs against an unchanged lockfile take a fast path, and pnpm shows a transient progress line while the registry round-trip runs.

  When fresh resolution picks an immature version, the behavior depends on minimumReleaseAgeStrict:

  • Loose mode — the default, in effect whenever minimumReleaseAge keeps its built-in 24-hour value — auto-adds the immature picks to minimumReleaseAgeExclude in pnpm-workspace.yaml and lets the install proceed. A single info message lists what was persisted.
  • Strict mode in an interactive terminal collects every immature direct AND transitive pick in one pass and prompts once with the full list. Approving adds them to minimumReleaseAgeExclude and the install continues; declining aborts before the lockfile, package.json, or node_modules is touched.
  • Strict mode in CI (or any non-TTY context) aborts with ERR_PNPM_NO_MATURE_MATCHING_VERSION listing every offending entry, instead of failing on the first one the resolver hit.

  minimumReleaseAgeStrict auto-enables whenever the user explicitly sets minimumReleaseAge (CLI flag, env var, global config.yaml, or pnpm-workspace.yaml); set minimumReleaseAgeStrict: false to keep loose-mode auto-collect even with an explicit minimumReleaseAge value. Closes #​10438, #​10488, #​11687.

• Allow redundant trailing base64 padding in .npmrc auth values and report invalid auth base64 with a pnpm error.

• Make pnpm self-update respect minimumReleaseAge (and minimumReleaseAgeExclude) when resolving which pnpm version to install.

  When the latest dist-tag points to a version newer than the configured age threshold, self-update now selects the newest mature version instead unless excluded by minimumReleaseAgeExclude.

  Also makes dlx and outdated surface invalid minimumReleaseAgeExclude patterns under the same ERR_PNPM_INVALID_MINIMUM_RELEASE_AGE_EXCLUDE error code already used by install, instead of leaking the internal ERR_PNPM_INVALID_VERSION_UNION / ERR_PNPM_NAME_PATTERN_IN_VERSION_UNION codes.

• Global installs respect global config build policy (e.g., dangerouslyAllowAllBuilds from config.yaml) when GVS is enabled #​9249.

  The global virtual-store (GVS) default allowBuilds = {} was applied before workspace manifest settings were read and before global config values (stripped by extractAndRemoveDependencyBuildOptions) were re-applied via globalDepsBuildConfig. This caused hasDependencyBuildOptions to return true (because {} is not null), blocking restoration of global config values like dangerouslyAllowAllBuilds. As a result, global installs skipped all build scripts even when the config explicitly allowed them.

  This fix moves the GVS default to after workspace manifest reading and globalDepsBuildConfig re-application, so that:

  1. Workspace manifest allowBuilds takes precedence (if present)
  2. Global config dangerouslyAllowAllBuilds is properly restored (if set and no workspace policy exists)
  3. Empty {} is only applied as a last resort when no policy is configured anywhere

• Honor --silent when verifyDepsBeforeRun: install auto-installs dependencies before pnpm run or pnpm exec, preventing install output from being written to stdout #​11636.

• Fix lockfile parsing failures when pnpm-lock.yaml contains CRLF line endings and multiple YAML documents #​11612.

• Anchor the side-effects-cache key and global-virtual-store hash to the project's script-runner Node — engines.runtime pin when present, shell node otherwise — instead of pnpm's own runtime.

  ENGINE_NAME (the <platform>;<arch>;node<major> prefix used as the side-effects-cache key and the engine portion of the GVS hash) was computed from process.version — the Node that runs pnpm itself. That was wrong in two situations:

  1. @pnpm/exe SEA bundle. The bundle has its own embedded Node, not the node on the user's PATH that actually spawns lifecycle scripts. Two pnpm installations on the same machine (one SEA, one npm-package) therefore disagreed on the cache key, partitioning the side-effects cache and the global virtual store across two Node majors even though both installs would run scripts on the same shell node.
  2. engines.runtime / devEngines.runtime pin. When a project pins a Node version via devEngines.runtime (pnpm v11+), pnpm downloads that Node into node_modules/node/ and uses it to run lifecycle scripts. But the hash still anchored to whichever Node ran pnpm itself, not to the pinned Node — so two installs of the same project with two different runner Nodes would still disagree on the GVS slot path even though scripts run on the same pinned Node.

  Three changes:

  • @pnpm/engine.runtime.system-node-version now exports engineName(nodeVersion?). Resolves the version in this order: explicit override → getSystemNodeVersion() (which already prefers node --version over process.version in SEA contexts) → process.version.
  • @pnpm/deps.graph-hasher now exports findRuntimeNodeVersion(snapshotKeys) — scans an iterable of lockfile snapshot keys for a node@runtime:<version> entry and returns its bare version string. calcDepState and calcGraphNodeHash/iterateHashedGraphNodes accept a nodeVersion? (in the options bag for the first, as a trailing parameter / ctx field for the others), forwarded to engineName(). The default (no override) preserves the pre-change behaviour. The legacy ENGINE_NAME constant in @pnpm/constants is unchanged so external consumers and existing tests keep working; in non-SEA, non-pinned contexts every value lines up.
  • Every install-side caller of the graph-hasher (@pnpm/installing.deps-resolver, @pnpm/installing.deps-restorer, @pnpm/installing.deps-installer, @pnpm/building.during-install, @pnpm/building.after-install, @pnpm/deps.graph-builder) now derives the project's pinned runtime via findRuntimeNodeVersion(Object.keys(graph)) once per invocation and threads it through.

  On upgrade, two one-time GVS slot churns are possible:

  • SEA-pnpm users without a runtime pin: slots that previously hashed under the embedded-Node major (e.g. node26) now hash under the shell-Node major (e.g. node24), matching what pacquet, the npm-published pnpm package, and any other pnpm-compatible tool already produce.
  • Projects with a devEngines.runtime pin: slots that previously hashed under the runner's Node major now hash under the pinned Node major, matching what the lifecycle scripts will actually run on.

  In both cases the old slots become prune-eligible.

• Resolve the GVS hash's engine portion per-snapshot when a dependency declares its own engines.runtime, instead of using an install-wide value.

  Pnpm's resolver desugars a dep's engines.runtime into dependencies.node: 'runtime:<version>', and the bin linker spawns that dep's lifecycle scripts through the pinned Node downloaded into <pkgDir>/node_modules/node/. The GVS hash and the side-effects-cache key prefix were still anchored to the install-wide runtime — so a pinning snapshot's slot encoded the wrong Node major, and a reinstall on the same host could read the cached side-effects under a key whose <platform>;<arch>;node<major> triple disagreed with the Node the build actually ran on.

  Per-snapshot resolution now matches what bins/linker already does on a per-package basis:

  • @pnpm/deps.graph-hasher adds readSnapshotRuntimePin(children) — reads the node entry from one snapshot's graph children and extracts the version from a node@runtime: value. Pairs with the existing findRuntimeNodeVersion(snapshotKeys) install-wide fallback (also now exported from @pnpm/deps.graph-hasher rather than @pnpm/engine.runtime.system-node-version, where it was a poor fit — system-node-version is about probing the host Node, not parsing lockfile-derived strings).
  • calcDepState and calcGraphNodeHash consult readSnapshotRuntimePin(graph[depPath].children) first and only fall back to the install-wide nodeVersion parameter when the snapshot doesn't pin its own Node.

  Pacquet mirrors the same precedence at the calc_graph_node_hash call site in package-manager/src/virtual_store_layout.rs — a new find_own_runtime_node_major(snapshot) helper reads each snapshot's dependencies for a node entry with Prefix::Runtime and overrides the install-wide engine when present.

  On upgrade, snapshots of dependencies that declare their own engines.runtime re-hash under that dep's pinned Node instead of the install-wide value. The old slots become prune-eligible. Closes #​11690.

• Fixed pnpm publish failing with a 404 when authentication relied on OIDC trusted publishing alongside an .npmrc written by actions/setup-node (_authToken=${NODE_AUTH_TOKEN}) without NODE_AUTH_TOKEN being set. Unresolved ${VAR} placeholders in auth values are now treated as empty rather than passed through verbatim, so the literal placeholder no longer surfaces as a bearer token when OIDC fallback is the intended auth source #​11513.

• Fix devEngines.packageManager (singular form, without onFail) defaulting to onFail: "error" instead of the documented pmOnFail: "download". As a result, a project that pinned a different pnpm version via devEngines.packageManager and ran pnpm install from a mismatched pnpm version failed with a hard error, even though the migration table from managePackageManagerVersions: true to pmOnFail: download (default) promises the install would auto-download the wanted version #​11676.

  The array form of devEngines.packageManager keeps its existing per-element defaults (error for the last entry, ignore for the rest), since those reflect explicit prioritization by the user. Explicit onFail values continue to win.

• Fix devEngines.packageManager not writing packageManagerDependencies to pnpm-lock.yaml when the lockfile lacks an env-doc entry. Previously the lockfile sync skipped resolution unless an existing packageManagerDependencies.pnpm entry needed refreshing, so a fresh install without onFail: "download" left the resolved pnpm version unrecorded — contradicting the documented behavior that the resolved version is stored in pnpm-lock.yaml #​11674.

• Warn when package.json contains a legacy pnpm field with settings pnpm no longer reads from package.json (e.g. pnpm.overrides, pnpm.patchedDependencies). Previously these were silently ignored after the upgrade from v10, leaving users unaware that their overrides/patched dependencies had stopped taking effect #​11677.

postcss/postcss (postcss)

v8.5.15

Compare Source

• Fixed declaration parsing performance (by @​homanp).

cloudflare/workers-sdk (wrangler)

v4.94.0

Compare Source

Minor Changes

• #​13897 52e9082 Thanks @​dario-piotrowicz! - Add automatic Cloudflare skills installation for AI coding agents

  Wrangler now detects AI coding agents and offers to install Cloudflare skill files from the cloudflare/skills GitHub repository. Users are prompted once interactively; subsequent runs skip the prompt. Use --install-skills to install without prompting.

• #​13989 f598eac Thanks @​MattieTK! - Print a QR code alongside the tunnel URL when sharing via Cloudflare Tunnel

  When a tunnel is started (via wrangler dev --tunnel or the Vite plugin with tunnel: true), a scannable QR code is now printed to the terminal beneath the tunnel URL. This makes it easy to open the tunnel on a mobile device without manually copying the URL.

  The QR code uses Unicode block characters for a compact representation and is generated best-effort -- if generation fails for any reason, the tunnel URL is still displayed as before.

• #​13467 3a1fbed Thanks @​deloreyj! - Add schedule property to Workflow bindings for cron-based triggering

  Note: This is a configuration-only change. Scheduled triggering of Workflow instances is not yet available — adding schedule to a Workflow binding will not result in scheduled invocations at this time. This change lays the groundwork for an upcoming feature.

  Workflow bindings in wrangler.json now accept an optional schedule field that configures one or more cron expressions to automatically trigger new workflow instances on a schedule.

  // wrangler.json
  {
    "workflows": [
      {
        "binding": "MY_WORKFLOW",
        "name": "my-workflow",
        "class_name": "MyWorkflow",
        "schedule": "0 9 * * 1"
      }
    ]
  }

  Multiple schedules can be provided as an array:

  {
    "workflows": [
      {
        "binding": "MY_WORKFLOW",
        "name": "my-workflow",
        "class_name": "MyWorkflow",
        "schedule": ["0 9 * * 1", "0 17 * * 5"]
      }
    ]
  }

  The schedule is sent to the Workflows control plane on wrangler deploy. Configuring schedule on a workflow binding that references an external script_name is an error — the schedule must be configured on the worker that defines the workflow.

Patch Changes

• #​13993 0733688 Thanks @​dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260520.1	1.20260521.1

• #​14008 fc1f7b9 Thanks @​petebacondarwin! - Fix Access Service Token authentication for applications that only allow service tokens

  When using remote bindings against a Worker behind a Cloudflare Access application configured to only allow Service Auth tokens (no interactive user authentication), Wrangler previously ignored the CLOUDFLARE_ACCESS_CLIENT_ID and CLOUDFLARE_ACCESS_CLIENT_SECRET environment variables and the request would fail with a 403.

  This happened because Wrangler detects Access by looking for a 302 redirect to cloudflareaccess.com. A service-auth-only Access application has no interactive login path, so it responds with a hard 403 instead of redirecting. Wrangler concluded the domain was not behind Access and skipped attaching the service token headers entirely.

  The env-var check now runs before the Access detection step, so the configured service token credentials are always used when present.

• #​12277 8c569c6 Thanks @​penalosa! - Include column names in D1 SQL export INSERT statements

  D1 SQL exports now include column names in INSERT statements (e.g., INSERT INTO "table" ("col1","col2") VALUES(...)). This ensures that exported SQL can be successfully imported even when the target table has columns in a different order than the original, which commonly occurs during iterative development when schemas evolve.

• Updated dependencies [0733688, 30657e1]:

  • miniflare@​4.20260521.0

v4.93.1

Compare Source

Patch Changes

• #​13978 fa1f61f Thanks @​sassyconsultingllc! - Bump ws from 8.18.0 to 8.20.1 to address GHSA-58qx-3vcg-4xpx

  GHSA-58qx-3vcg-4xpx / CVE-2026-45736 reports an uninitialized-memory disclosure in ws@<8.20.1 when a TypedArray is passed as the reason argument to WebSocket.close(). The fix shipped in ws@8.20.1 on 2026-05-12. This change bumps the workspace catalog entry so that miniflare, wrangler, and @cloudflare/vite-plugin all pick up the patched release.

• #​13977 2679e05 Thanks @​dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260518.1	1.20260519.1

• #​13984 7e40d98 Thanks @​dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260519.1	1.20260520.1

• #​13963 adc9221 Thanks @​gabivlj! - Preserve sibling container image tags during local dev cleanup

  Wrangler now keeps other cloudflare-dev image tags from the same dev session when multiple containers share a Dockerfile. Previously, duplicate-image cleanup could remove earlier container tags if Docker BuildKit produced the same image ID for each build.

• #​13839 735852d Thanks @​matingathani! - fix: show actionable hint when /memberships returns a bad-credentials error (code 9106)

  Previously, wrangler threw a raw Cloudflare API error ("Missing X-Auth-Key, X-Auth-Email or Authorization headers") with no guidance. Now it emits a UserError explaining that an environment variable such as CLOUDFLARE_API_TOKEN, CLOUDFLARE_API_KEY, or CLOUDFLARE_EMAIL may be set to an invalid value, and suggests running wrangler logout / wrangler login to re-authenticate.

• #​13912 d803737 Thanks @​petebacondarwin! - Fix /cdn-cgi/* host validation incorrectly accepting subdomains of exact configured routes

  Miniflare's /cdn-cgi/* host/origin validator was treating exact configured routes the same as wildcard configured routes, so a request whose Host or Origin hostname was a subdomain of an exact route (e.g. sub.my-custom-site.com for a my-custom-site.com/* route) was incorrectly accepted. Exact configured routes and the configured upstream hostname are now required to match the request hostname exactly. Subdomain matching is only applied to wildcard routes such as *.example.com/*. Localhost hostnames continue to be allowed as before.

  This affects wrangler dev and local development through @cloudflare/vite-plugin, both of which use Miniflare under the hood.

• #​13919 c7eab7f Thanks @​petebacondarwin! - Fix the outbound CF-Worker header reflecting the route pattern hostname instead of the parent zone, and falling back to <worker-name>.example.com under vite dev, vitest-pool-workers, and getPlatformProxy

  Two related issues affected the CF-Worker header on outbound subrequests in local development:

  1. Under @cloudflare/vite-plugin, @cloudflare/vitest-pool-workers, and getPlatformProxy, the header fell back to <worker-name>.example.com even when routes were configured, because unstable_getMiniflareWorkerOptions and the equivalent getPlatformProxy worker-options path did not propagate a zone value to Miniflare. This broke local development against services that reject unknown CF-Worker hosts (for example, Apple WeatherKit returns 403 Forbidden).
  2. Across the above paths and wrangler dev --local, when a route used the zone_name field (for example { pattern: "foo.example.com/*", zone_name: "example.com" }), the header was set to the pattern's hostname (foo.example.com) rather than the zone name (example.com). Production sets CF-Worker to the zone name that owns the Worker, so this was inconsistent with deployed behaviour.

  Both bugs are fixed: the new unstable_getMiniflareWorkerOptions / getPlatformProxy path now propagates a zone derived from the first configured route, and all four local-dev paths now prefer a route's explicit zone_name over the pattern hostname when computing that zone. When zone_name isn't set, the existing best-effort behaviour is preserved — for wrangler dev this means dev.host is still honoured as a local override and the pattern hostname is used as a final fallback. Resolving the parent zone for zone_id-only, custom_domain, or plain-string routes would require an API lookup, so locally we still approximate it with the pattern hostname.

  Note: dev.host is intentionally not consulted by the unstable_getMiniflareWorkerOptions / getPlatformProxy paths — the dev config block is specific to wrangler dev.

• #​13990 e04e180 Thanks @​petebacondarwin! - Improve the log message shown when an asset upload attempt fails and is retried

  The retry message now reports which attempt is being made (e.g. Asset upload failed. Retrying... 1 of 5 attempts.), making it easier to gauge how close Wrangler is to exhausting its retry budget. The raw error object is no longer appended to this user-facing message; it is instead logged at debug level (visible via WRANGLER_LOG=debug).

• #​13954 62abf97 Thanks @​petebacondarwin! - Read the on-disk OAuth state lazily so CLOUDFLARE_API_TOKEN from .env takes priority correctly

  Wrangler previously read its OAuth state from the user auth config file (for example ~/.config/.wrangler/config/default.toml) eagerly at module-import time. That happens before .env files are loaded, so the in-memory state would always hold the OAuth tokens even when the user only wanted to authenticate via CLOUDFLARE_API_TOKEN. If that stored OAuth token happened to be expired, Wrangler would try to refresh it (and fail), aborting the command with Failed to fetch auth token: 400 Bad Request and Not logged in. — even though a valid API token was in scope.

  Wrangler now reads the auth config file on demand, after .env has been loaded. When CLOUDFLARE_API_TOKEN (or CLOUDFLARE_API_KEY + CLOUDFLARE_EMAIL) is present, the OAuth state on disk is no longer consulted, the OAuth refresh endpoint is no longer called, and the env-based token is used directly. Sibling-process refresh-token rotation is also handled naturally because every check reads the current file contents.

  Internally, the exported reinitialiseAuthTokens() function is removed — there is no module-level OAuth cache left to invalidate.

  Fixes #​13744.

• #​13951 e349fe0 Thanks @​sejoker! - Enforce minimum 60 second interval for R2 Data Catalog sinks

  R2 Data Catalog sinks now require a minimum --roll-interval of 60 seconds to prevent compaction issues in the R2 Data Catalog. This validation is applied when creating sinks via wrangler pipelines sinks create with type r2-data-catalog, and during the interactive wrangler pipelines setup flow.

  Regular R2 sinks are not affected and can still use intervals as low as 10 seconds.

• #​13959 da0fa8c Thanks @​dmmulroy! - Recognize Artifacts repositories that are still being created

  Wrangler's Artifacts repo status type now accepts the creating lifecycle state alongside existing in-progress statuses.

• #​13964 a5c9365 Thanks @​danielrs! - Use dedicated API endpoint for wrangler secret bulk

  wrangler secret bulk now uses a more efficient, dedicated API endpoint. This reduces the operation from 2 API calls to 1 and eliminates the risk of accidentally affecting non-secret bindings.

• Updated dependencies [fa1f61f, 2679e05, 7e40d98, d803737, 59cd880, e8c2031]:

  • miniflare@​4.20260520.0

v4.93.0

Compare Source

Minor Changes

• #​13901 aac7ca0 Thanks @​bghira! - Add wrangler ai models schema command for fetching model schemas

  You can now run wrangler ai models schema <model> to fetch the input and output schema for a Workers AI model from the public model catalog schema endpoint.

• #​12656 ae047ee Thanks @​mikenomitch! - Add --containers-rollout=none

  This allows you to skip deploying a container. This is useful if you know that your container is not going to be updated or you don't have Docker locally, but still want to make changes to your Worker.

• #​13901 aac7ca0 Thanks @​bghira! - Add wrangler ai models list command for querying the Workers AI model catalog

  wrangler ai models list accepts --search, --task, --author, --source, and --hide-experimental, matching the public model catalog search endpoint.

Patch Changes

• #​13948 b25dc0d Thanks @​dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260515.1	1.20260518.1

• #​13882 a4f22bc Thanks @​matingathani! - Throw a clear error when a D1 migration is cancelled instead of silently returning

• #​13950 f78d435 Thanks @​dario-piotrowicz! - Improve the Docker CLI error message to be more actionable.

  Include a link to Docker installation docs, platform-specific instructions for starting the daemon, and guidance for alternative Docker-compatible CLIs.

• #​11896 c5c9e20 Thanks @​staticpayload! - Surface remote proxy session errors

  When remote bindings fail to start, include the controller reason and root cause in the error message to make failures like missing cloudflared clearer.

• #​13932 ebf4b24 Thanks @​zebp! - Fix local Workflow startup when compatibility flags include experimental

  Miniflare now deduplicates compatibility flags for the internal Workflow engine service. This prevents wrangler dev from failing with Compatibility flag specified multiple times: experimental when the user's Worker already enables that flag.

• #​13929 895baf5 Thanks @​Caio-Nogueira! - Prompt to provision a workers.dev subdomain before deploying Workflows

  Wrangler now checks for the account-level workers.dev subdomain when deploying Workflows, even if the Worker is not being published to workers.dev. If the subdomain has not been registered yet, Wrangler prompts to create one before calling the Workflows deploy API so users avoid an opaque server-side deployment failure.

• #​13930 7bcdf45 Thanks @​shiminshen! - Sweep stale .wrangler/tmp/* dirs left behind by abnormal exits

  A wrangler dev session creates .wrangler/tmp/bundle-* and .wrangler/tmp/dev-* directories at startup and removes them via a signal-exit hook on graceful shutdown. When the process exited abnormally (SIGKILL, OOM, host crash) those directories were left behind and accumulated across sessions, slowing down dependency-walking tools that follow the bundle-emitted absolute-path imports.

  wrangler now sweeps entries in .wrangler/tmp/ older than 24 hours when a new temporary directory is requested, bounding the leak regardless of how prior sessions exited.

• Updated dependencies [b25dc0d, ebf4b24, b27eb18]:

  • miniflare@​4.20260518.0

v4.92.0

Compare Source

Minor Changes

• #​13670 506aa02 Thanks @​elithrar! - Add wrangler artifacts commands for managing Artifacts repos and repo tokens.

  This adds CLI support for the Artifacts control-plane workflows that were previously only available through the API. You can now list and inspect namespaces, create, list, inspect, and delete repos, and issue repo-scoped tokens when you need to authenticate git access.

  The new commands support both human-readable output and --json output so they fit existing Wrangler automation patterns.

• #​13916 be8a98c Thanks @​emily-shen! - Add --keep-vars flag to wrangler versions upload, matching the existing behavior in wrangler deploy. When set, environment variables configured via the dashboard are preserved rather than being deleted before the upload.

Patch Changes

• #​13926 19ed49a Thanks @​dependabot! - Update dependencies of "miniflare", "wrangler"

  The following dependency versions have been updated:

  Dependency	From	To
  workerd	1.20260511.1	1.20260515.1

• #​11471 3ff0a50 Thanks @​HW13! - Improve wrangler types --env-interface for multi-worker projects.

  Custom env interfaces generated by wrangler types no longer expand from Cloudflare.Env, avoiding some unintended type expansion when multiple workers' generated types are used together.

• #​13910 bf688f7 Thanks @​timoconnellaus! - Fix Failed to fetch auth token: 401 Unauthorized from sibling-rotated refresh tokens

  refreshToken previously used the refresh token from module-level localState, which is populated once at startup and never re-read. OAuth refresh tokens are single-use, so when a sibling wrangler process (in another repo, another shell, or a parallel script) refres

✂ Note

PR body was truncated to here.

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • Only on Monday (* * * * 1)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.