@agent-native/core-corpus 0.206.0-nightly-20261009124314 → 0.206.0

package/corpus/README.md

@@ -32,4 +32,4 @@ rg -n "defineAction|useActionQuery" node_modules/@agent-native/core-corpus/corpu
 
 ## Generated Counts
 
-- template files: 11666
+- template files: 11704

package/corpus/templates/analytics/.agents/skills/onboarding-journeys/SKILL.md

@@ -2,9 +2,10 @@
 name: onboarding-journeys
 description: >-
   Build the onboarding journey tree (signup to first output, with % splits,
-  last-observed-step counts, and example replays) and render a screenshot per
-  step. Use when asked for an onboarding storyboard, where users stop being
-  observed, or frames for each onboarding step.
+  last-observed-step counts, window-bounded later activity, and example
+  replays) and render a screenshot per step. Use when asked for an onboarding
+  storyboard, each cohort's last recorded step or later activity within a
+  window, or frames for each onboarding step.
 scope: dev
 ---
 
@@ -21,7 +22,9 @@ The pipeline runs without a UI. Codex does all four steps from the terminal:
 3. Call Design's `create-journey-canvas` with
    `{ title, tree, frames: [{ nodeKey, exampleIndex, imageUrl?, attachmentRef?, width, height, capturedAt }], designId? }`
    -> `{ designId, url, nodeCount, frameCount }`. Build `frames` from
-   `frames/manifest.json`.
+   `frames/manifest.json`. Design reads event timestamps and replay offsets from
+   `tree`; keep `sourceEventAt`, `replayAt`, and `capturedAt` in the capture
+   manifest.
 4. Lay out and review the canvas.
 
 ## The tree
@@ -34,7 +37,8 @@ window, `app`, `maxDepth`, `minNodeSessions` (small branches merge into an
 ```ts
 type JourneyExample = { sessionId: string; recordingId: string | null; ts: string; offsetMs: number | null; viewport: { width: number; height: number } | null; viewportReason?: string; replayUrl?: string };
 type JourneyNode = { key: string; label: string; parentKey: string | null; depth: number; kind: "step" | "other"; n: number; pctOfRoot: number; pctOfParent: number; dropoffN: number; dropoffPct: number; deeperN: number; examples: JourneyExample[] };
-type JourneyTree = { window: { from: string; to: string }; app: string; rootN: number; coverage: { sessionsWithEvents: number; sessionsWithReplay: number; truncated: boolean }; nodes: JourneyNode[]; standaloneSetup?: { rootN: number; coverage: { sessionsWithEvents: number; sessionsWithReplay: number; truncated: boolean }; nodes: JourneyNode[] }; notes?: string[] };
+type JourneyFollowup = { status: "complete" | "incomplete"; incompleteReason?: "journey_event_read_truncated" | "journey_event_read_invalid" | "journey_event_read_may_have_shifted" | "terminal_cohort_query_too_large" | "followup_aggregate_truncated" | "followup_aggregate_invalid" | "terminal_cohort_mismatch"; observationCutoff: string; observationFollowupDurationMs: { min: number; max: number; mean: number } | null; rightCensoredAtWindowEnd: true; coverage: { journeyEventRead: { rows: number; pages: number; truncated: boolean; paginationConsistency: "stable" | "may_have_shifted" }; followupAggregateRead: { rows: number | null; queries: number; truncated: boolean }; cohortSessions: number | null }; laterRecordedActivityWithinWindow: { total: number | null; byTerminalStepKey: Record<string, number> | null }; noLaterRecordedActivityWithinWindow: { total: number | null; byTerminalStepKey: Record<string, number> | null } };
+type JourneyTree = { window: { from: string; to: string }; app: string; rootN: number; coverage: { sessionsWithEvents: number; sessionsWithReplay: number; truncated: boolean }; nodes: JourneyNode[]; followUp: JourneyFollowup; standaloneSetup?: { rootN: number; coverage: { sessionsWithEvents: number; sessionsWithReplay: number; truncated: boolean }; nodes: JourneyNode[] }; notes?: string[] };
 ```
 
 Analytics returns cohort nodes with counts. When extending a Design storyboard
@@ -63,6 +67,28 @@ top-level tree to `journey:capture`; the capture CLI reads top-level `nodes`.
   exit: a blocked tracker or an event outside the window reads the same.
   `deeperN` counts sessions with a later observed step that is not represented
   as a child, including paths past `maxDepth` or a node-list cap.
+- `followUp` groups onboarding sessions by the key of their terminal selected
+  step. `laterRecordedActivityWithinWindow` counts sessions with any later
+  native Analytics event in the same session; `noLaterRecordedActivityWithinWindow`
+  counts the remaining sessions. The read applies the same authenticated
+  user/org scope, date window, app, identity bridge, test exclusion, and
+  Builder.io email filter as the journey read. It uses one
+  frozen `observationCutoff` for every event page and the single aggregate
+  query. `observationFollowupDurationMs`
+  summarizes the time from each terminal selected step to that cutoff.
+  `rightCensoredAtWindowEnd: true` means the no-later count is right-censored:
+  no later event was recorded before the cutoff; this is not an abandonment or
+  churn outcome. The aggregate contains counts only and does not expose the
+  session IDs or member identity keys used internally. The consistency field
+  reports `may_have_shifted` when the event read uses multiple `OFFSET` pages;
+  a late-arriving event can change page membership inside a historical window
+  too. If either read truncates, page boundaries may have shifted, or the
+  terminal cohort cannot be covered in one query under the 800,000-character
+  SQL limit or 50,000-token parser limit, `status` is `incomplete` and all new
+  cohort counts and follow-up duration are `null`; `incompleteReason`
+  identifies the limiting read. Do not report percentages from that partial
+  result. Existing journey counts and denominators remain independent of this
+  follow-up read.
 - `maxDepth` defaults to 8 and is bounded at 40. Request `maxDepth: 40` for a
   deeper pass. When sessions continue past the requested depth,
   `coverage.truncated` is true and the boundary node's `deeperN` says how many
@@ -114,6 +140,19 @@ and `manifest.json` with `frames`, `failures` (explicit, with a reason), and
 bounded batches only through the requested offsets and validates every fetched
 chunk. Fix failures or report them; do not paint over a missing frame.
 
+`sourceEventAt` preserves the JourneyTree example's original event timestamp;
+`replayAt` is `startedAt + offsetMs`, and `capturedAt` is when the CLI rendered
+the PNG. Every visible iframe must have a corresponding recorded child document
+by the requested replay time. If the child document is missing, the frame is
+listed as a failure with code `replay_iframe_content_unavailable` and counts in
+`diagnostics`. If unsupported 3D projection, rounded ancestor clipping, CSS
+clip paths, masks, or visibility-altering filters prevent the audit from
+verifying visibility, the frame is listed with code
+`replay_iframe_visibility_unverifiable`; do not treat uncertainty as either
+missing content or a successful audit. The check covers every visible iframe
+because replay can omit its original source attribute while rebuilding an
+isolated frame.
+
 Authenticate to the deployed app, never a local database: run
 `npx -y @agent-native/core@latest connect https://analytics.agent-native.com --client codex`
 (the CLI reads the bearer it writes to `~/.codex/config.toml`), or pass
@@ -126,6 +165,9 @@ remote images and fonts are not fetched. The output manifest marks this as
 `remoteAssets: "not-fetched"`. Keep recorded URLs and CSS intact for rrweb
 playback; network controls belong at the capture boundary. `--upload` stores
 PNGs through the private upload action, which checks recording access again.
+Design's own sandboxed template previews opt into the cooperative iframe
+recorder, so their child DOM can be replayed without granting the parent access
+to the preview document or credentials.
 
 Viewport comes from the recording's first rrweb Meta event, stored by replay
 ingest in `session_recordings.metadata.viewport` (`first` and `last`). Older

package/corpus/templates/analytics/CHANGELOG.md

@@ -3,6 +3,20 @@
 All notable user-facing changes to Agent-Native Analytics are documented here. Open it any
 time from the command menu (Cmd+K → "What's new") or from Settings.
 
+## 2026-10-09
+
+### Improved
+
+- Onboarding journeys include Builder connection and custom-key outcomes through deeper paths
+- Onboarding journeys now show later activity within a chosen window, mark no-later results as right-censored, and identify when follow-up counts are incomplete.
+- You can add app origins to a public key's replay allowlist without replacing its existing origins.
+
+### Fixed
+
+- Journey captures now flag visible previews missing from the recording instead of saving blank screenshots
+- Journey recordings flag iframe content when clipping, masks, or filters make its visibility uncertain.
+- Large onboarding journeys keep their step counts when follow-up coverage exceeds query limits
+
 ## 2026-10-08
 
 ### Added
@@ -19,6 +33,9 @@ time from the command menu (Cmd+K → "What's new") or from Settings.
 
 ### Fixed
 
+- Analytics journey paths group short resource IDs consistently across apps
+- Analytics keeps onboarding counts complete when only standalone setup events exceed the journey read limit.
+- Onboarding journey trees show Home chat setup choices in a separately counted tree and group equivalent artifact screens.
 - Agent-shared recording links now load across batched playback chunks
 - Fixed onboarding journey queries against BigQuery.
 - Session replay agent links can read their recording manifest and playback chunks.

package/corpus/templates/analytics/actions/get-onboarding-journey.ts

@@ -34,7 +34,7 @@ const isoDate = z
 
 export default defineAction({
   description:
-    'Build the onboarding journey tree for a date window: per-session step sequences (signup, onboarding questions, setup method, app entry, first action, first output) folded into a prefix tree. format "tree" returns { window, app, rootN, coverage, nodes } for sessions that entered onboarding, plus standaloneSetup for Home chat integration events in sessions without onboarding; its root and percentages have a separate denominator. Each node has n, pctOfRoot, pctOfParent, dropoffN/dropoffPct (last observed step, not a confirmed exit), deeperN (sessions with a later observed step omitted below this node), and example sessions with recordingId, a recording-start-relative offsetMs, and viewport. format "summary" returns the same counts as indented outlines with no examples. Check each tree\'s coverage.truncated first: true means its event sample, requested depth, or node list cut that tree; notes explain why. Use it to plan onboarding storyboards or find where users stop. For a standalone storyboard, pass standaloneSetup as the top-level tree to journey:capture.',
+    'Build the onboarding journey tree for a date window: per-session step sequences (signup, onboarding questions, setup method, app entry, first action, first output) folded into a prefix tree. format "tree" returns { window, app, rootN, coverage, nodes } for sessions that entered onboarding, plus standaloneSetup for Home chat integration events in sessions without onboarding; its root and percentages have a separate denominator. Each node has n, pctOfRoot, pctOfParent, dropoffN/dropoffPct (last observed step, not a confirmed exit), deeperN (sessions with a later observed step omitted below this node), and example sessions with recordingId, a recording-start-relative offsetMs, and viewport. Both formats include followUp counts of sessions with and without any later native Analytics event in the same session after each terminal selected step, observed before one frozen observationCutoff. followUp.rightCensoredAtWindowEnd is true: no later event means none was recorded within that bounded window, not abandonment or churn. If the event read may be incomplete, the aggregate query truncates or exceeds the SQL size or parser-token limit, or cohort coverage mismatches, followUp.status is incomplete, incompleteReason names the cause, and its cohort counts and duration are null. format "summary" returns the same tree counts as indented outlines with no examples. Check each tree\'s coverage.truncated first: true means its event sample, requested depth, or node list cut that tree; notes explain why. Use it to plan onboarding storyboards or see each cohort\'s last recorded selected step. For a standalone storyboard, pass standaloneSetup as the top-level tree to journey:capture.',
   schema: z.object({
     dateFrom: isoDate.describe(
       "Inclusive UTC start date, YYYY-MM-DD. Sessions that began earlier appear mid-journey, so start a day before the period you care about.",

package/corpus/templates/analytics/app/pages/sessions/SessionReplayStoryboardExportDialog.tsx

@@ -185,6 +185,7 @@ export function SessionReplayStoryboardExportDialog({
 
   function captureErrorMessage(error: unknown): string {
     if (error instanceof ReplayScreenshotAssetError) {
+      console.warn("Replay screenshot capture rejected", error.reason);
       return t("sessions.screenshotUnsupportedAssets");
     }
     return error instanceof Error

package/corpus/templates/analytics/app/pages/sessions/session-replay-screenshot.ts

@@ -1,5 +1,15 @@
 export class ReplayScreenshotAssetError extends Error {
-  constructor() {
+  constructor(
+    readonly reason:
+      | "unsupportedAsset"
+      | "fontReadiness"
+      | "cloneDocument"
+      | "duplicateCloneMarker"
+      | "cloneRoot"
+      | "cloneElement"
+      | "cloneStageFrame"
+      | "clonePseudoElement" = "unsupportedAsset",
+  ) {
     super("Replay contains media or images that cannot be captured safely");
     this.name = "ReplayScreenshotAssetError";
   }
@@ -914,7 +924,7 @@ export function inlineReplayAssets(
   const originalDocuments = [originalDocument];
   const clonedDocuments = markedReplayDocuments(clonedDocument, captureId);
   if (clonedDocuments.length !== 1) {
-    throw new ReplayScreenshotAssetError();
+    throw new ReplayScreenshotAssetError("cloneDocument");
   }
 
   const clonesByMarker = new Map<string, Element>();
@@ -922,7 +932,9 @@ export function inlineReplayAssets(
     for (const element of renderedElements(cloned)) {
       const marker = element.getAttribute(REPLAY_SCREENSHOT_MARKER);
       if (!marker?.startsWith(`${captureId}-`)) continue;
-      if (clonesByMarker.has(marker)) throw new ReplayScreenshotAssetError();
+      if (clonesByMarker.has(marker)) {
+        throw new ReplayScreenshotAssetError("duplicateCloneMarker");
+      }
       clonesByMarker.set(marker, element);
     }
   }
@@ -948,7 +960,7 @@ export function inlineReplayAssets(
         rootMarker,
     );
     if (!cloned || rootMarker !== `${captureId}-${documentIndex}-0`) {
-      throw new ReplayScreenshotAssetError();
+      throw new ReplayScreenshotAssetError("cloneRoot");
     }
     const documentAssets = assets.get(original) ?? new Map();
 
@@ -966,7 +978,7 @@ export function inlineReplayAssets(
           originalElement.tagName !== clonedElement.tagName &&
           !originalElement.localName.includes("-"))
       ) {
-        throw new ReplayScreenshotAssetError();
+        throw new ReplayScreenshotAssetError("cloneElement");
       }
       if (originalElement instanceof view.HTMLIFrameElement) {
         const screenshot = iframeScreenshots.get(originalElement);
@@ -1059,7 +1071,9 @@ export function inlineReplayAssets(
             : "___html2canvas___pseudoelement_after";
         if (!clonedElement.classList.contains(generatedClass)) continue;
         const pseudoElement = pseudoElements[pseudoElementIndex++];
-        if (!pseudoElement) throw new ReplayScreenshotAssetError();
+        if (!pseudoElement) {
+          throw new ReplayScreenshotAssetError("clonePseudoElement");
+        }
         if (
           pseudoStyles.display === "none" ||
           pseudoStyles.visibility === "hidden" ||
@@ -1119,7 +1133,7 @@ async function assertDocumentFontsReady(document: Document): Promise<void> {
       fontSet.ready,
       new Promise<never>((_resolve, reject) => {
         timeoutId = window.setTimeout(
-          () => reject(new ReplayScreenshotAssetError()),
+          () => reject(new ReplayScreenshotAssetError("fontReadiness")),
           REPLAY_FONT_TIMEOUT_MS,
         );
       }),
@@ -1339,7 +1353,9 @@ export async function captureReplayScreenshot(
         const clonedFrame = clonedStageRoot.querySelector(
           `iframe[${REPLAY_SCREENSHOT_MARKER}="${stageFrameMarker}"]`,
         );
-        if (!clonedFrame) throw new ReplayScreenshotAssetError();
+        if (!clonedFrame) {
+          throw new ReplayScreenshotAssetError("cloneStageFrame");
+        }
         const image = clonedStageRoot.ownerDocument.createElement("img");
         image.alt = "";
         image.src = replayImageUrl;

package/corpus/templates/analytics/changelog/2026-10-09-journey-captures-now-flag-visible-previews-missing-from-the-.md

@@ -0,0 +1,6 @@
+---
+type: fixed
+date: 2026-10-09
+---
+
+Journey captures now flag visible previews missing from the recording instead of saving blank screenshots

package/corpus/templates/analytics/changelog/2026-10-09-journey-recordings-flag-iframe-content-that-css-clipping-or-.md

@@ -0,0 +1,6 @@
+---
+type: fixed
+date: 2026-10-09
+---
+
+Journey recordings flag iframe content when clipping, masks, or filters make its visibility uncertain.

package/corpus/templates/analytics/changelog/2026-10-09-large-onboarding-journeys-keep-their-step-counts-when-follow.md

@@ -0,0 +1,6 @@
+---
+type: fixed
+date: 2026-10-09
+---
+
+Large onboarding journeys keep their step counts when follow-up coverage exceeds query limits

package/corpus/templates/analytics/changelog/2026-10-09-onboarding-journeys-now-count-later-activity-within-a-frozen.md

@@ -0,0 +1,6 @@
+---
+type: improved
+date: 2026-10-09
+---
+
+Onboarding journeys now show later activity within a chosen window, mark no-later results as right-censored, and identify when follow-up counts are incomplete.

package/corpus/templates/analytics/package.json

@@ -8,6 +8,7 @@
     "build": "agent-native build && tsx scripts/emit-netlify-dashboard-report-cron.ts",
     "start": "agent-native start",
     "test": "vitest --run",
+    "test:bigquery-emulator": "vitest --run --config vitest.bigquery.config.ts",
     "format.fix": "oxfmt --write .",
     "typecheck": "agent-native typecheck",
     "migrate:production": "tsx scripts/migrate-production.ts",