@@ -111,6 +111,8 @@ trusted policy and reassess changed submission evidence; neither example grants
 
 Do not automatically close a PR if a maintainer in the trusted `TRIAGE_EXEMPTIONS.td` list has
 commented or submitted a review, including on earlier heads, or if `triage:keep-open` is present.
+Comments that triage itself posted, such as an earlier notice or question, are not maintainer
+involvement, even when they were posted from a maintainer's account on that maintainer's behalf.
 Leave protected PRs open for maintainer decision. This does not grant an author exemption,
 eligibility, or review approval.
 
@@ -126,7 +128,17 @@ comment text but do not write to GitHub. Keep the eligibility finding separate f
   This is a routing exception, not a claim that the PR passed eligibility or correctness review.
 - **Closure warranted.** The assessment is complete and establishes a specific policy violation.
   Prepare a clear explanation under [closure and reconsideration](../../../CONTRIBUTING.md#closure-and-reconsideration),
-  post it, and close the PR. Verify that the explanation is present and the PR is closed. The comment
+  post it, then read the repository variable `PR_LIMITS_ENFORCE`
+  (`gh variable get PR_LIMITS_ENFORCE -R pingdotgg/t3code`). When it prints `true`, apply
+  `triage:closing` as the last thing you do to the PR, verify that the explanation and the label
+  are present, and do not close the PR yourself:
+  [pr-limits.yml](../../../.github/workflows/pr-limits.yml) closes it 72 hours after the label. Any
+  activity on the PR after the label, including a further comment or label from you, removes the
+  label; treat an author's reply or push as new evidence, reassess, and apply the label again only
+  if the finding still holds. When the command prints any other value, or fails saying the variable
+  was not found, the variable is off and nothing would close a labeled PR: close it yourself and
+  verify that the explanation is present and the PR is closed. Only another failure, such as a
+  permission error, means the variable cannot be read; then report the closure as incomplete. The comment
   must name the violated rule, cite supporting submission evidence, link the maintained guide section,
   and give a concrete remedy. Missing contributor evidence can warrant closure; explain what must be
   established. Missing product approval calls for maintainer discussion, not an agent's product decision.
@@ -173,9 +185,10 @@ description, evidence, approvals, existing triage comments, PR open/closed state
 label when relevant. Reassess changes that could invalidate the finding. Reuse an existing explanation
 only if it still matches the current finding; avoid duplicate comments, closures, or label applications.
 
-For closure, establish that the required explanation is posted before closing. If posting fails or its
-result is ambiguous, read back the comments before retrying or proceeding. If the comment succeeds but
-closure fails, preserve the comment and retry only the unfinished closure after checking current state.
+For closure, establish that the required explanation is posted before applying `triage:closing` or
+closing. If posting fails or its result is ambiguous, read back the comments before retrying or
+proceeding. If the comment succeeds but the label or the closure fails, preserve the comment and retry
+only that unfinished step after checking current state.
 Handle a failed or ambiguous label application the same way: inspect labels before any retry. If access
 or state still cannot be verified, report an incomplete action and the step needed to resume. Do not
 retry blindly or claim success for an unverified action.
 
@@ -5,10 +5,10 @@ description: Test T3 Code's web and desktop UI through its built-in Browser pane
 
 # Test T3 web and desktop
 
-Use T3's built-in Browser panel for verification. If its tools are absent or
-the panel reports unavailable, explain the blocker and stop verification.
-Do not install or switch to another automation system. For native mobile
-testing, use [test-t3-mobile](../test-t3-mobile/SKILL.md).
+Use T3's built-in Browser panel for verification. Follow the recovery steps
+below before declaring it unavailable. Do not install or switch to another
+automation system. For native mobile testing, use
+[test-t3-mobile](../test-t3-mobile/SKILL.md).
 
 ## Start the app
 
@@ -36,8 +36,13 @@ model, such as Haiku for Claude or Luna for Codex.
 
 ## Use the Browser panel
 
-Call `preview_status`, then `preview_open` if the Browser panel is
-closed. Navigate to the complete startup pairing URL once with
+Call `preview_status`. If no tab can be automated, call `preview_open`, including
+when status reports `available: false` or `tabId: null`. Status can report
+unavailable simply because no tab exists yet; opening a tab needs no user action.
+Stop verification and explain the blocker only if the tools are absent or
+`preview_open` reports unavailable or fails.
+
+Navigate to the complete startup pairing URL once with
 `preview_navigate`, then use `preview_snapshot` and T3's interaction tools.
 If the token was consumed or expired, run `node apps/server/src/bin.ts pair`
 for a fresh one. Keep using the same tab.
 
@@ -36,8 +36,9 @@ vp run mobile:install <ios|android> <device-id>
 
 This installs the dev client for the checkout's native fingerprint: from the
 device, the local cache, or the shared cache in seconds, building once only
-when no machine has that fingerprint yet (5-15 minutes, logged to the file it
-names). The device id may be omitted when exactly one device is booted. It
+when no machine has that fingerprint yet: in the `Mobile Dev Client` workflow
+on Blacksmith when HEAD is pushed and `gh` may dispatch it (the command waits),
+otherwise here (5-15 minutes, logged to the file it names). The device id may be omitted when exactly one device is booted. It
 prints one JSON line; on failure it names the next command. Mobile verification
 includes that step unless the user prohibits it.
 
 
@@ -1,7 +1,7 @@
 # Security policy
 
 Report security vulnerabilities affecting T3 Code or T3 Tools-operated infrastructure to
-[security@ping.gg](mailto:security@ping.gg). Please do not disclose them publicly until we have had
+[security@t3.codes](mailto:security@t3.codes). Please do not disclose them publicly until we have had
 a reasonable opportunity to investigate and remediate them.
 
 See the [full security policy](https://t3.codes/security-policy) for reporting details, scope,
 
@@ -5,6 +5,8 @@ outputs:
     value: ${{ steps.detect.outputs.typecheck }}
   build:
     value: ${{ steps.detect.outputs.build }}
+  desktop_launch:
+    value: ${{ steps.detect.outputs.desktop_launch }}
   test:
     value: ${{ steps.detect.outputs.test }}
   test_packages:
 
@@ -13,6 +13,10 @@ inputs:
     description: Arguments for `vp install`, such as filters. Empty installs the whole workspace.
     required: false
     default: ""
+  save-pnpm-store:
+    description: Set to "false" when a concurrent job with the same install args saves the pnpm store
+    required: false
+    default: "true"
 runs:
   using: composite
   steps:
@@ -123,10 +127,14 @@ runs:
         key: ${{ steps.pnpm_verified.outputs.cache-primary-key }}
 
     # A filtered install that started from the current full store adds nothing
-    # worth a second copy of it.
+    # worth a second copy of it. A stacked PR, based on a branch other than the
+    # default, saves a ~1 GB store only its own reruns can read, so it only restores.
+    # Concurrent jobs with the same install args reserve one key, so the losers
+    # compress the store for nothing; all but one pass save-pnpm-store: false.
     - name: Save pnpm store
       if: >-
-        inputs.install == 'true' &&
+        inputs.install == 'true' && inputs.save-pnpm-store == 'true' &&
+        (github.event_name != 'pull_request' || github.base_ref == github.event.repository.default_branch) &&
         steps.pnpm_cache.outputs.cache-hit != 'true' &&
         (steps.pnpm_store.outputs.namespace == 'full' ||
           steps.pnpm_cache.outputs.cache-matched-key != format('pnpm-full-v1-{0}-{1}-{2}', runner.os, runner.arch, hashFiles('pnpm-lock.yaml', 'pnpm-workspace.yaml', 'package.json')))
 
@@ -40,42 +40,67 @@ async function assertReleaseSource({ github, context, releaseChannel }) {
 
 const isNightlyTag = (tag) => /^v.*-nightly\./.test(tag) || tag.startsWith("nightly-v");
 
-async function findLatestNightly({ github, context }) {
-  const releases = await github.paginate(github.rest.repos.listReleases, {
-    ...context.repo,
-    per_page: 100,
-  });
-  return releases
-    .filter((release) => !release.draft && release.published_at && isNightlyTag(release.tag_name))
-    .sort((a, b) => Date.parse(b.published_at) - Date.parse(a.published_at))[0];
+// Only the three fields the gate reads. The REST list also returns every asset,
+// which made a full scan tens of megabytes.
+const RELEASES_QUERY = `
+  query ($owner: String!, $repo: String!, $cursor: String) {
+    repository(owner: $owner, name: $repo) {
+      releases(first: 100, after: $cursor, orderBy: { field: CREATED_AT, direction: DESC }) {
+        pageInfo { hasNextPage endCursor }
+        nodes { tagName isDraft publishedAt }
+      }
+    }
+  }
+`;
+
+// Returns the nightly with the newest publishedAt. Release order follows the
+// tagged commit date, not publication, so every page is read unless `isSettled`
+// says the newest nightly seen so far already decides the caller's answer.
+async function findLatestNightly({ github, context, isSettled = () => false }) {
+  let latest;
+  let cursor = null;
+  do {
+    const { repository } = await github.graphql(RELEASES_QUERY, { ...context.repo, cursor });
+    const { nodes, pageInfo } = repository.releases;
+    for (const release of nodes) {
+      if (release.isDraft || !release.publishedAt || !isNightlyTag(release.tagName)) continue;
+      if (!latest || Date.parse(release.publishedAt) > Date.parse(latest.publishedAt)) {
+        latest = release;
+      }
+    }
+    if (latest && isSettled(latest)) return latest;
+    cursor = pageInfo.hasNextPage ? pageInfo.endCursor : null;
+  } while (cursor);
+  return latest;
 }
 
 async function shouldReleaseNightly({ github, context, core, now = Date.now() }) {
-  const lastNightly = await findLatestNightly({ github, context });
+  const isInGap = (release) => now - Date.parse(release.publishedAt) < MINIMUM_RELEASE_GAP_MS;
+  // Any nightly inside the gap means the newest one is too, so stop reading.
+  const lastNightly = await findLatestNightly({ github, context, isSettled: isInGap });
 
   if (!lastNightly) {
     core.info("No published nightly found. Proceeding with release.");
     return true;
   }
 
-  if (now - Date.parse(lastNightly.published_at) < MINIMUM_RELEASE_GAP_MS) {
-    core.info(`Nightly ${lastNightly.tag_name} was published less than six hours ago. Skipping.`);
+  const tag = lastNightly.tagName;
+  if (isInGap(lastNightly)) {
+    core.info(`Nightly ${tag} was published less than six hours ago. Skipping.`);
     return false;
   }
 
   const { data: comparison } = await github.rest.repos.compareCommitsWithBasehead({
     ...context.repo,
-    basehead: `${lastNightly.tag_name}...${context.sha}`,
+    basehead: `${tag}...${context.sha}`,
     per_page: 1,
   });
   if (comparison.status !== "ahead") {
-    core.info(
-      `Candidate commit is ${comparison.status} relative to ${lastNightly.tag_name}. Skipping.`,
-    );
+    core.info(`Candidate commit is ${comparison.status} relative to ${tag}. Skipping.`);
     return false;
   }
 
-  core.info(`New commits since ${lastNightly.tag_name}, and the six-hour gap has passed.`);
+  core.info(`New commits since ${tag}, and the six-hour gap has passed.`);
   return true;
 }
 
@@ -85,7 +110,7 @@ async function resolveLatestNightlyCommit({ github, context, core }) {
     throw new Error("No published nightly found. Stable releases build the latest nightly commit.");
   }
 
-  const tag = lastNightly.tag_name;
+  const tag = lastNightly.tagName;
   const { data: commit } = await github.rest.repos.getCommit({ ...context.repo, ref: tag });
   const version = /^(?:nightly-)?v(\d+\.\d+\.\d+)-nightly\./.exec(tag)?.[1];
   if (!version) {