Update Dex Skills to 0.62.0 for server-zd v0.1.15 (#1176)

Author: longquanzheng

skills/.claude-plugin/marketplace.json

@@ -11,7 +11,7 @@
       "name": "superdurable-dex",
       "source": "./",
       "description": "App Builder-led Dex development in a repository-backed coding workspace.",
-      "version": "0.61.0",
+      "version": "0.62.0",
       "skills": [
         "./dex-app-builder",
         "./dex-sdk",

skills/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "superdurable-dex",
-  "version": "0.61.0",
+  "version": "0.62.0",
   "description": "Build Dex process products in a repository-backed coding workspace.",
   "author": {
     "name": "Super Durable",

skills/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "superdurable-dex",
-  "version": "0.61.0",
+  "version": "0.62.0",
   "description": "Build Dex process products in a repository-backed coding workspace.",
   "author": {
     "name": "Super Durable",

skills/.cursor-plugin/marketplace.json

@@ -11,7 +11,7 @@
       "name": "superdurable-dex",
       "source": "./",
       "description": "App Builder-led Dex development in a repository-backed coding workspace.",
-      "version": "0.61.0",
+      "version": "0.62.0",
       "logo": "assets/logo.png"
     }
   ]

skills/.cursor-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "superdurable-dex",
-  "version": "0.61.0",
+  "version": "0.62.0",
   "description": "Build Dex process products in a repository-backed coding workspace.",
   "author": {
     "name": "Super Durable"

skills/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## 0.62.0 - 2026-11-01
+
+- Document server-zd v0.1.15 in Core operations. `standby.mode: hot` runs extra processes with the same node ID; a hot standby follows the log from object storage and takes the lease within about one lease poll after it becomes available, while a crash still waits for the old lease to expire. The default `cold` is unchanged. A standby answers public API calls with UNAVAILABLE, so it needs a TCP or exec startup probe and liveness probe, and the `dex.HotStandby` health name reports whether a hot standby is ready. Record the new hot standby and takeover metrics, and that every process must run v0.1.15 before any configuration adds a `standby` section.
+- Replace the v0.1.14 AdminService groundwork wording: server-zd v0.1.15 serves AdminService. `dexcli admin step-down --yes` makes the active process give up its lease and keep running as a standby, and `dexcli admin describe` shows the active process. Admin calls need a JWT whose audience is in `adminAudiences` of its issuer's `jwt` authenticator; static tokens and API key exchange never authorize them. Retry an unknown-outcome step down only with the node ID and node epoch that the failed call reported. Record the fix for stale reads after lease loss, which also affected cold standbys, and known issue #1174, where a process that takes the lease back after a step down can exit and needs a restart. Link the operation guide instead of copying it.
+- Advance `DEX_SERVER_BASELINE` to server-zd/v0.1.15 and `DEX_CLI_BASELINE` to cli-v1.33.0 (which bundles it), and refresh the pinned CLI and server-zd source links. cli-v1.33.0 changes only the CLI, so the SDK baselines stay at v1.32.0 (Rust at sdk-rust/v1.31.0) and `DEX_WEB_BASELINE` stays at web/v1.32.0; the Web and protocol trees are unchanged. Synchronize the bundle and plugin manifests at 0.62.0.
+- Check the v0.1.15 hot standby, takeover, step down, admin credential, and metric statements against the tag in `check-upstream-baselines.py`, drop the v0.1.14 statements that said AdminService is not served, check the new Core operations statements, and check the latest server-zd release at server-zd/v0.1.15. Add mutation tests.
+
 ## 0.61.0 - 2026-11-01
 
 - Document server-zd v0.1.14 in Core operations. The ParadeDB and PostgreSQL search backends now project through a durable outbox written in the same local commit as each search document change, so a database outage is lag, never lost data, and the in-memory queue, overflow-triggered partition rebuilds, and startup catch-up rebuild are gone. Describe the initial load and its logged reasons (`no_progress`, `no_outbox_progress`, `enabled_after_progress`, `acked_after_progress`, and `ahead_of_committed_state`), the one-time reload of every partition on the first start after the upgrade, during which SearchFlows for a loading partition waits or returns UNAVAILABLE, the read-your-writes wait that now also covers restarts, takeovers, and initial loads, the INTERNAL and fenced UNAVAILABLE stopped states, the optional `search.outboxBatchSize`, `search.outboxPollInterval`, `search.outboxAckInterval`, and `search.outboxMaxAckEntries` settings with their defaults, and the outbox backlog, lag, initial-load, and skipped-partition metrics. Core read-after-write keeps search eventual.

skills/DEX_CLI_BASELINE

@@ -1 +1 @@
-cli-v1.32.0
+cli-v1.33.0

skills/DEX_SERVER_BASELINE

@@ -1 +1 @@
-server-zd/v0.1.14
+server-zd/v0.1.15

skills/VERSION

@@ -1 +1 @@
-0.61.0
+0.62.0

skills/dex-app-builder/references/dex-web-v2.md

@@ -1,6 +1,6 @@
 # Dex Web v2 and FDG 2.0
 
-Reference capability baselines: Dex Web `web/v1.32.0`, `cli-v1.32.0`, the
+Reference capability baselines: Dex Web `web/v1.32.0`, `cli-v1.33.0`, the
 server-zd release it bundles, and the SDK source baseline in
 [bundle-baselines.md](../../dex-sdk/references/core/bundle-baselines.md).
 These validate the guidance, not an instruction to upgrade an application.