@agent-native/core 0.136.0 → 0.136.3

package/corpus/README.md

@@ -31,4 +31,4 @@ rg -n "defineAction|useActionQuery" node_modules/@agent-native/core/corpus
 
 ## Generated Counts
 
-- template files: 7499
+- template files: 7515

package/corpus/templates/analytics/.agents/skills/analysis-workspace/SKILL.md

@@ -48,6 +48,28 @@ window, then read them back selectively for synthesis.
   `timed_out`, chunk the work or persist intermediate progress with
   `workspaceWrite` and re-run.
 
+## Delegated Analysis Handoff
+
+Background sub-agents start with a clean context. A handoff for a large
+analysis must include the exact staged dataset id or workspace path, the
+records/chunk range to process, the output path, and the checkpoint or resume
+operation. The child must read that staged input before making provider calls.
+
+- If a readable staged file or dataset already exists, process it in place. Do
+  not re-page the provider endpoint to recreate the same corpus.
+- If the named staged input is missing or unreadable, stop with an explicit
+  missing-input error. Do not silently fall back to a full provider sweep.
+- For more than 500 Gong records, do not use `gong-calls(exhaustive=true)`, a
+  per-call transcript loop, or a single `provider-api-request` `fetchAllPages`
+  call as the analysis itself. If the term is a configured keyword tracker,
+  stage `/calls/extensive` tracker results and reduce them with a Data Program
+  or `query-staged-dataset`. Otherwise use `provider-corpus-job` with a staged
+  call-id input and continue raw transcript retrieval in bounded, checkpointed
+  batches.
+- Write a compact result or progress marker after every completed chunk. A
+  final answer written only after the whole corpus is processed is not a
+  checkpoint.
+
 ## Workspace File Helpers
 
 Inside `run-code`, use the workspace helper functions:
@@ -105,11 +127,15 @@ narrow, where records must be joined across systems, or where absence matters:
 
 1. Discover the provider surface with `provider-api-catalog` and
    `provider-api-docs` when endpoint/filter/pagination details are uncertain.
-2. Pull the relevant records with `provider-api-request`. Use `fetchAllPages`
-   for cursor pagination, `stageAs` for queryable staged datasets, and
-   `saveToFile` for large raw responses.
-3. Use `run-code` to call `providerFetch` or `appAction` in loops, write
-   intermediate files, normalize records, join identity fields, and aggregate.
+2. For a broad or absence-sensitive corpus, start with the provider's native
+   search or indexed tracker surface when one exists. Otherwise use
+   `provider-api-request` as the raw ingestion step with `stageAs` or
+   `saveToFile`, then reduce the staged data with `query-staged-dataset` or a
+   Data Program. Use `provider-corpus-job` for durable paginated or batched
+   transcript/body scans so each page or batch is checkpointed durably.
+3. Use `run-code` to read staged data, write intermediate files, normalize
+   records, join identity fields, and aggregate. For long code, use
+   `background: true` and persist progress after each chunk.
 4. Validate coverage before synthesis. Track pages fetched, records inspected,
    truncation flags, aborted calls, and records skipped for missing joins.
 5. Finalize with the answer plus coverage and caveats. If coverage is partial,

package/corpus/templates/analytics/.agents/skills/dashboard-management/SKILL.md

@@ -111,6 +111,13 @@ When the user asks for a dashboard:
 6. Persist with `update-dashboard`, not raw SQL or settings writes.
 7. Navigate to it with `pnpm action navigate --view=adhoc --dashboardId=<id>`.
 
+An explicit dashboard request authorizes the complete non-destructive build in
+the same turn. After querying or scaffolding, continue through extension-data
+seeding/refresh, dashboard save/embed, and navigation; do not ask whether to
+proceed or leave an empty Custom Block shell. Ask only when metric scope/grain
+is materially ambiguous, the change is destructive, or it has an external side
+effect such as sending email or outreach.
+
 Layout is always **1 column when the available content width is below the `md` threshold** (panels stack), then expands to the configured column count at/above it. The grid uses a container query, so it also stacks when the agent sidebar narrows the content pane — not only at narrow viewports. So picking 3 or 4 columns is fine — the renderer keeps narrow layouts readable automatically.
 
 ```bash

package/corpus/templates/analytics/.agents/skills/gong/SKILL.md

@@ -24,9 +24,18 @@ dive that asks what happened in customer conversations.
 - `account-deep-dive` — first choice for named account/deal deep dives that
   need HubSpot plus Gong. It searches by account/deal/company/contact domain,
   loads Gong call details, and returns compact transcript excerpts for synthesis.
-- `gong-calls` — list recent calls, search by company/domain/person/email, fetch
-  a single transcript by call ID, or return transcript excerpts for matching
-  calls.
+- `gong-calls` — bounded convenience reads: list recent calls, search by
+  company/domain/person/email, fetch a single transcript by call ID, or return
+  transcript excerpts for a small matching cohort. Its `transcriptQuery` is a
+  local scan after batched transcript retrieval, not Gong's server-side
+  keyword search.
+
+For broad or absence-sensitive Gong work, prefer the raw provider API surface:
+use `provider-api-catalog` / `provider-api-docs`, then stage `/calls/extensive`
+or another raw endpoint with `provider-api-request`. Reduce staged tracker or
+call data with `query-staged-dataset` or a Data Program. Use
+`provider-corpus-job` only when raw transcript bodies are required and no
+native/indexed search can answer the question.
 
 ## Route Synthesis and Evidence Separately
 
@@ -97,10 +106,12 @@ For account or deal deep dives:
    quoting, debugging, or export.
 6. Ground qualitative findings in the transcript excerpts and state how many
    calls were inspected.
-7. Page through calls using `cursor`/offset when you need broad coverage — for
-   large accounts with many calls, do NOT stop at the first page. For very large
-   pulls (100+ calls), prefer chunked background processing rather than a single
-   blocking fetch.
+7. Page through calls using `cursor`/offset when you need bounded account
+   coverage — do NOT stop at the first page. For broad scans, do not use
+   `gong-calls(exhaustive=true)` or one blocking `provider-api-request` call as
+   the analysis. Stage the raw response and reduce it with a Data Program or
+   `query-staged-dataset`; use `provider-corpus-job` for checkpointed transcript
+   batches when raw bodies are necessary.
 
 Example:
 
@@ -113,6 +124,9 @@ emails, and domains through `/calls/extensive`. Treat call details and transcrip
 excerpts as evidence; treat missing coverage as a gap, not proof that the topic
 never came up.
 
+Title/party matching is discovery only. Never use a title pre-filter as proof
+that a phrase is absent from transcripts.
+
 If transcript loading fails for a call, report that gap instead of inferring the
 conversation content from title, date, or participants.
 
@@ -127,28 +141,46 @@ this cohort mention X?" — where missing a single call makes the answer wrong
 NOT conclude absence from a sampled `includeTranscripts` result. Choose the
 smallest complete path for the bounded cohort:
 
-1. **Named account, up to 50 calls, qualitative review.** Make one `gong-calls`
+1. **Configured keyword tracker.** If the requested term is already a Gong
+   keyword tracker exposed in API results, use `provider-api-request` against
+   `/calls/extensive` with `contentSelector.exposedFields.content.trackers`.
+   Stage the paginated calls, then flatten and filter tracker ids, names, or
+   phrases with `query-staged-dataset` or a Data Program. This is the preferred
+   broad keyword path because Gong has already indexed the tracker. It does not
+   answer arbitrary terms that are not configured as trackers.
+
+2. **Named account, up to 50 calls, qualitative review.** Make one `gong-calls`
    call with `exhaustive=true`, a bounded `after`/`before` window, and
    `includeTranscripts=true`. The action discovers every matching call and
    fetches transcript excerpts in batches, returning `transcriptCoverage`. Do
    not fetch the same call IDs one at a time afterward.
 
-2. **Bounded mention/absence search, up to 200 calls.** Make one `gong-calls`
+3. **Bounded arbitrary mention/absence search, up to 200 calls.** Make one
+   `gong-calls`
    call with `exhaustive=true`, the date window, `transcriptQuery`, and an
-   adequate `transcriptScanLimit`. Report its coverage fields and only make an
-   absence claim when coverage is complete.
+   adequate `transcriptScanLimit`. This is a bounded local scan after batched
+   transcript retrieval. Report its coverage fields and only make an absence
+   claim when coverage is complete.
 
-3. **Larger or multi-account corpus.** Prefer `provider-corpus-job` with
-   `mode: "batch-search"` over one-call-at-a-time loops. Use
+4. **Larger or multi-account arbitrary corpus.** Use
    `provider-api-catalog(provider: "gong")` and its `corpusRecipes` if you need
-   the exact shape. The canonical request is `POST /calls/transcript` with
+   the exact shape. Stage call ids with the raw API, then use
+   `provider-corpus-job` with `mode: "batch-search"` and 20-call transcript
+   batches. The canonical request is `POST /calls/transcript` with
    `batch.itemBodyPath: "filter.callIds"`, `batch.responseItemsPath:
 "callTranscripts"`, `batch.batchSize: 20`, `search.textPaths:
 ["transcript"]`, and `search.idPaths: ["callId"]`. Feed the staged/discovered
-   call IDs through `batch.inputDatasetId` + `batch.inputValuePath` or through
-   `batch.items`.
+   call ids through `batch.inputDatasetId` + `batch.inputValuePath` or through
+   `batch.items`. This is a durable full scan, not a server-side keyword query;
+   do not implement it as a per-call loop in `run-code` or a delegated agent.
+
+When delegating a corpus chunk to an Agent Teams sub-agent, include the exact
+staged dataset id or `scratch/...` path and the output/checkpoint path in the
+task brief. The child starts with a clean context and must read the staged
+input before making any provider request. If that input is unavailable, it
+must report the missing input rather than re-fetching the entire Gong corpus.
 
-4. **Use `run-code` only for joins/reductions around the corpus path.** After
+5. **Use `run-code` only for joins/reductions around the corpus path.** After
    the transcript job exists, use `run-code`, `query-staged-dataset`, or job
    results to join hits back to deals/accounts, compute variants, dedupe, and
    format evidence. A `run-code` loop over `gong-calls(transcript: id)` is a
@@ -166,5 +198,6 @@ matches found. Never turn "I inspected a sample" into "no call mentions X".
 | `transcriptMaxChars`     | 8 000   | 100 000 |
 
 For large account deep-dives or competitive analyses spanning many calls, use
-`limit: 50-200` and `transcriptLimit: 20-50`. For very large datasets, the
-`provider-api-request` escape hatch can page through the full Gong call list.
+`limit: 50-200` and `transcriptLimit: 20-50`. For very large datasets, use the
+checkpointed `provider-corpus-job` path rather than paging the full Gong call
+list inside one agent action.

package/corpus/templates/analytics/.agents/skills/provider-api/SKILL.md

@@ -54,6 +54,24 @@ It is not required for hosted Agent Native provider access. Do not install or
 vendor that plugin by default; its public repository currently declares no
 license.
 
+## Gong
+
+Gong's UI has indexed Words or phrases search, but the public REST endpoints
+used here do not expose an arbitrary transcript-text filter. The efficient API
+path is a configured keyword tracker: request `content.trackers` from
+`POST /calls/extensive`, stage the paginated calls, and flatten/filter tracker
+hits with `query-staged-dataset` or a Data Program. This only works for terms
+already configured as trackers and exposed in API results.
+
+For an arbitrary term, use Gong native Search when that connected surface is
+available. If raw transcript evidence is required, stage call IDs with the raw
+API and use `provider-corpus-job`'s 20-call transcript batches. Do not loop over
+`gong-calls(transcript: id)` from `run-code` or a delegated agent. A corpus job
+improves durability and checkpointing, but it still scans transcript bodies and
+should be described as an expensive fallback. For recurring tracker-based
+reports, save the staged fetch and reduction as a Data Program instead of
+creating a new provider-specific action.
+
 ## Workflow
 
 1. Use a first-class action when it exactly fits the request.
@@ -61,7 +79,9 @@ license.
    shape, or pagination mode, switch to `provider-api-catalog` for that
    provider. Check `corpusRecipes` first when the user asks for broad body-text
    searches across transcripts, messages, tickets, issues, notes, documents, or
-   conversation logs.
+   conversation logs. For Gong, use the configured keyword-tracker recipe when
+   it covers the requested term; otherwise use the raw transcript recipe only
+   when the provider's native search surface is unavailable or insufficient.
 3. If the endpoint or payload is not obvious, use `provider-api-docs` to fetch
    the official docs/spec URL from the catalog.
 4. Call `provider-api-request` with the exact provider method, path, query, and
@@ -72,8 +92,10 @@ license.
      charge history), **add `stageAs`** to avoid context-window truncation.
    - For multi-page results, **also add `pagination`** config to fetch all pages
      server-side in one call (cursor / page / offset modes supported).
-5. After staging, call `query-staged-dataset` to aggregate. Only the compact
-   summary (counts, sums, sample rows) needs to flow into the context window.
+5. After staging, call `query-staged-dataset` to aggregate, or save a Data
+   Program when the provider pull should become a cached, refreshable source.
+   Only the compact summary (counts, sums, sample rows) needs to flow into the
+   context window. Treat the raw request as ingestion, not the final analysis.
 6. For source-record body searches, use the raw body endpoint or native search
    endpoint for that record type. Parent/container metadata such as call lists,
    channel lists, ticket titles, summaries, or briefs is discovery evidence, not

package/corpus/templates/analytics/AGENTS.md

@@ -7,7 +7,7 @@ artifact; legacy analyses remain readable only for compatibility.
 **This agent is Claude with BigQuery/Gong/HubSpot/Prometheus access.** Reason
 over fetched data here; do not suggest another AI tool.
 
-Prompt cap: ~6,000 characters; put detail in `.agents/skills/*`.
+Prompt cap: 6,000; put detail in `.agents/skills/*`.
 
 Before building common workspace or agent UI, read `agent-native-toolkit`; use
 `customizing-agent-native` for the configure → compose → eject → propose
@@ -37,9 +37,6 @@ ladder.
    threads). Chunking a question a query could answer is the most expensive
    mistake available here. See `adhoc-analysis`.
 
-Use `ask-question` at most once when an ambiguous metric, range, or grain would
-change the numbers.
-
 ## Core Rules
 
 - A sibling app asking over A2A sends a question or shaped input, never SQL. It
@@ -55,9 +52,11 @@ change the numbers.
   filters, row count/sample size, join method, caveats.
 - Use actions for sources, queries, charts, dashboards, and sharing. Don't bypass
   access checks with raw SQL for ownable resources.
-- Provider actions are shortcuts, not limits — escalate to
-  `provider-api-catalog` / `provider-api-docs` / `provider-api-request` when a
-  canned action is too narrow. See `provider-api`.
+- Provider actions are bounded shortcuts, not limits. For broad or
+  absence-sensitive Gong work, stage raw API data and use
+  `query-staged-dataset` or a Data Program; if raw transcript bodies are
+  required, use `provider-corpus-job`. See `provider-api`, `data-programs`, and
+  `gong`.
 - Create dashboards, panels, or saved artifacts only when explicitly asked;
   suggest and wait otherwise. Scope them to the question, avoid decorative
   metrics, and never modify existing dashboards without a directive.

package/corpus/templates/analytics/actions/gong-calls.ts

@@ -485,7 +485,7 @@ export default defineAction({
   // model round trips while still staying well below the model context limit.
   maxResultChars: 100_000,
   description:
-    "Query Gong sales calls, transcripts, and users. Pass --users for user list, --transcript for one transcript, --company to search by company/domain/person/email. For bounded account-level transcript mention/search questions, set transcriptQuery to search matching transcripts server-side and return coverage counts plus snippets instead of large transcript blobs. For deal, customer, objection, next-step, or deep-dive analysis, set includeTranscripts=true only when you need broad qualitative context rather than a specific term search. For an 'all calls' transcript review in a bounded account/date window, use exhaustive=true with after/before and includeTranscripts=true; the action batches transcript retrieval and reports complete/partial coverage instead of forcing one tool call per transcript. For very broad cohorts or defensible absence searches, use provider-api-catalog(provider='gong') and provider-corpus-job mode='batch-search' against /calls/transcript after call-id discovery.",
+    "Query bounded Gong sales-call evidence. Pass --users for the user list, --transcript for one transcript, or --company for a bounded search by company/domain/person/email. For account-level transcript mention questions, transcriptQuery performs a case-insensitive local scan after batched transcript retrieval and returns coverage counts plus snippets; it is not Gong's server-side keyword search. Use includeTranscripts=true for bounded qualitative context. For broad keyword, tracker, cross-account, or absence-sensitive work, use provider-api-catalog/provider-api-docs, stage the raw Gong API response, then use query-staged-dataset or a Data Program; use provider-corpus-job only when raw transcript bodies are required.",
   schema: z.object({
     users: cliBoolean.optional().describe("Set to true to list Gong users"),
     transcript: z.string().optional().describe("Call ID to get transcript"),
@@ -538,7 +538,7 @@ export default defineAction({
       .string()
       .optional()
       .describe(
-        "Case-insensitive phrase to search inside matching call transcripts. Use this for bounded account/call searches where the matching set is already small. For broad cohort or exhaustive absence research, stage Gong calls/transcripts through provider-api-request and run-code instead. Returns coverage counts and short snippets only, not full transcripts.",
+        "Case-insensitive phrase to search inside matching call transcripts after batched retrieval. Use only for bounded account/call searches where the matching set is already small. For broad cohort or exhaustive absence research, use the raw Gong API tracker/staging path plus query-staged-dataset or a Data Program, or provider-corpus-job when raw transcript bodies are required. Returns coverage counts and short snippets only, not full transcripts.",
       ),
     transcriptScanLimit: z.coerce
       .number()
@@ -552,7 +552,7 @@ export default defineAction({
     exhaustive: cliBoolean
       .optional()
       .describe(
-        "Return EVERY matching call in the window instead of stopping at `limit` — use this for complete cohort/account coverage when 'how many' or absence matters. By default this is metadata-only. With includeTranscripts=true, transcripts are fetched in batches up to transcriptLimit (all matched calls by default when there are 50 or fewer) and explicit coverage is returned. Always bound it with after/before or a small days window: an unbounded exhaustive scan pages the whole Gong org and can hit the function timeout.",
+        "Return EVERY matching call in the window instead of stopping at `limit` — use this only for bounded account/cohort coverage. By default this is metadata-only. With includeTranscripts=true, transcripts are fetched in batches and explicit coverage is returned. For broad or absence-sensitive work, prefer the raw Gong API tracker/staging path plus query-staged-dataset or a Data Program; use provider-corpus-job for durable raw-transcript scans. Always bound this with after/before or a small days window.",
       ),
     after: z
       .string()

package/corpus/templates/analytics/actions/provider-api-catalog.ts

@@ -20,6 +20,6 @@ export default createProviderApiCatalogAction(
     }),
     http: { method: "GET" },
     guidance:
-      "First-class provider actions in this app are convenience shortcuts, not capability limits. When an action cannot express the needed endpoint, object type, filter, request body, pagination mode, API version, or source-record body coverage, inspect docs/spec URLs and corpusRecipes here, then call provider-api-request or provider-corpus-job with the exact provider API method/path/query/body.",
+      "First-class provider actions in this app are bounded convenience shortcuts, not capability limits. When an action cannot express the needed endpoint, object type, filter, request body, pagination mode, API version, or source-record body coverage, inspect docs/spec URLs and corpusRecipes here. Use provider-api-request as the raw ingestion/staging step, then query-staged-dataset or save-data-program for broad searches, joins, and reusable aggregates; use provider-corpus-job for durable paginated or batched body scans.",
   },
 );

package/corpus/templates/analytics/actions/query-agent-native-analytics.ts

@@ -15,7 +15,7 @@ function resolveScope() {
 
 export default defineAction({
   description:
-    "Query the built-in first-party Analytics source: events recorded through this app's analytics collector endpoint (/track) and session replay summaries recorded through /api/analytics/replay. This source does not require an external provider connection. Use it for questions about app/site traffic, product events, template/app usage, conversions, session recordings, LLM/agent observability, model cost, token volume, latency, and other first-party data collected by this analytics app. Use source-specific actions such as BigQuery, GA4, Mixpanel, PostHog, or Amplitude when the user asks for those sources or the relevant data lives there. SQL may read analytics_events and session_recordings only; session_replay_chunks is intentionally unavailable, and reads are automatically scoped to the current user/org. analytics_events columns include event_name, timestamp, event_date, user_id, anonymous_id, user_key, session_id, app, template, signed_in, url, path, hostname, referrer, properties, and context. LLM observability events use event_name = '$ai_generation'; useful properties include $ai_trace_id/run_id, $ai_session_id/thread_id, $ai_model/model, $ai_provider/provider, $ai_input_tokens/input_tokens, $ai_output_tokens/output_tokens, cache_read_tokens, cache_write_tokens, cost_cents_x100, $ai_total_cost_usd/cost_usd, duration_ms/$ai_latency, status, tool_calls, successful_tools, failed_tools, tools, tools_truncated, delegated, delegation_protocol, caller_app, a2a_task_id, parent_run_id, parent_turn_id, and error_message/$ai_error. The bounded tools array contains names, relative start times, durations, statuses, and coarse error classes only, never args or results; failed runs and interrupted tools remain queryable. session_recordings columns include id, session_id, user_id, anonymous_id, user_key, started_at, ended_at, duration_ms, chunk_count, event_count, page_count, error_count, rage_click_count, app, template, status, first_url, last_url, path, hostname, referrer, and metadata.",
+    "Query the built-in first-party Analytics source: events recorded through this app's analytics collector endpoint (/track) and session replay summaries recorded through /api/analytics/replay. This source does not require an external provider connection. Use it for questions about app/site traffic, product events, template/app usage, conversions, session recordings, LLM/agent observability, model cost, token volume, latency, and other first-party data collected by this analytics app. Use source-specific actions such as BigQuery, GA4, Mixpanel, PostHog, or Amplitude when the user asks for those sources or the relevant data lives there. SQL may read analytics_events and session_recordings only; session_replay_chunks is intentionally unavailable, and reads are automatically scoped to the current user/org. analytics_events columns include event_name, timestamp, event_date, user_id, anonymous_id, user_key, session_id, app, template, signed_in, url, path, hostname, referrer, properties, and context. LLM observability events use event_name = '$ai_generation'; useful properties include $ai_trace_id/run_id, $ai_session_id/thread_id, $ai_model/model, $ai_provider/provider, $ai_input_tokens/input_tokens, $ai_output_tokens/output_tokens, cache_read_tokens, cache_write_tokens, cost_cents_x100, $ai_total_cost_usd/cost_usd, duration_ms/$ai_latency, status, tool_calls, successful_tools, failed_tools, tools, tools_truncated, delegated, delegation_protocol, caller_app, delegation_task_id, a2a_task_id, parent_run_id, parent_turn_id, and error_message/$ai_error. Agent Teams child runs use delegation_protocol = 'agent-team' and retain their own run_id while linking to the launching run through parent_run_id. The bounded tools array contains names, relative start times, durations, statuses, and coarse error classes only, never args or results; failed runs and interrupted tools remain queryable. session_recordings columns include id, session_id, user_id, anonymous_id, user_key, started_at, ended_at, duration_ms, chunk_count, event_count, page_count, error_count, rage_click_count, app, template, status, first_url, last_url, path, hostname, referrer, and metadata.",
   schema: z.object({
     sql: z
       .string()

package/corpus/templates/analytics/app/i18n-data.ts

@@ -194,6 +194,10 @@ const enUS = {
     openDashboardTemplates: "Open catalog",
     languageTitle: "Language",
     languageLabel: "Interface language",
+    errorEmailNotifications: "Email new error alerts",
+    errorEmailNotificationsDescription:
+      "Send an email when a new JavaScript error is captured. Off by default.",
+    errorEmailNotificationsSaveFailed: "Couldn't save the email preference.",
     about: "About",
     aboutDescription:
       "Analytics is a tool for connecting data sources and building custom dashboards. Connect Google Analytics, BigQuery, Stripe, and more, then ask the agent to create dashboards.",
@@ -4253,6 +4257,10 @@ export const messagesByLocale = {
       openDashboardTemplates: "打开目录",
       languageTitle: "语言",
       languageLabel: "界面语言",
+      errorEmailNotifications: "通过电子邮件接收新的错误提醒",
+      errorEmailNotificationsDescription:
+        "捕获新的 JavaScript 错误时发送电子邮件。默认关闭。",
+      errorEmailNotificationsSaveFailed: "无法保存电子邮件偏好设置。",
       about: "关于",
       aboutDescription:
         "Analytics 用于连接数据源并构建自定义仪表板。连接 Google Analytics、BigQuery、Stripe 等，然后让代理创建仪表板。",
@@ -4471,6 +4479,11 @@ export const messagesByLocale = {
       openDashboardTemplates: "Abrir catálogo",
       languageTitle: "Idioma",
       languageLabel: "Idioma de la interfaz",
+      errorEmailNotifications: "Recibir nuevas alertas de errores por email",
+      errorEmailNotificationsDescription:
+        "Envía un email cuando se capture un nuevo error de JavaScript. Desactivado de forma predeterminada.",
+      errorEmailNotificationsSaveFailed:
+        "No se pudo guardar la preferencia de email.",
       about: "Acerca de",
       aboutDescription:
         "Analytics conecta fuentes de datos y crea paneles personalizados. Conecta Google Analytics, BigQuery, Stripe y más, y pide al agente que cree paneles.",
@@ -4693,6 +4706,12 @@ export const messagesByLocale = {
       openDashboardTemplates: "Ouvrir le catalogue",
       languageTitle: "Langue",
       languageLabel: "Langue de l'interface",
+      errorEmailNotifications:
+        "Recevoir les nouvelles alertes d’erreur par e-mail",
+      errorEmailNotificationsDescription:
+        "Envoyer un e-mail lorsqu’une nouvelle erreur JavaScript est capturée. Désactivé par défaut.",
+      errorEmailNotificationsSaveFailed:
+        "Impossible d’enregistrer la préférence e-mail.",
       about: "À propos",
       aboutDescription:
         "Analytics connecte des sources de données et crée des tableaux de bord personnalisés. Connectez Google Analytics, BigQuery, Stripe et plus encore, puis demandez à l'agent de créer des tableaux de bord.",
@@ -4922,6 +4941,11 @@ export const messagesByLocale = {
       openDashboardTemplates: "Katalog öffnen",
       languageTitle: "Sprache",
       languageLabel: "Oberflächensprache",
+      errorEmailNotifications: "Neue Fehlerwarnungen per E-Mail erhalten",
+      errorEmailNotificationsDescription:
+        "Eine E-Mail senden, wenn ein neuer JavaScript-Fehler erfasst wird. Standardmäßig deaktiviert.",
+      errorEmailNotificationsSaveFailed:
+        "Die E-Mail-Einstellung konnte nicht gespeichert werden.",
       about: "Info",
       aboutDescription:
         "Analytics verbindet Datenquellen und erstellt benutzerdefinierte Dashboards. Verbinde Google Analytics, BigQuery, Stripe und mehr und bitte den Agenten, Dashboards zu erstellen.",
@@ -5142,6 +5166,10 @@ export const messagesByLocale = {
       openDashboardTemplates: "カタログを開く",
       languageTitle: "言語",
       languageLabel: "インターフェース言語",
+      errorEmailNotifications: "新しいエラーアラートをメールで受け取る",
+      errorEmailNotificationsDescription:
+        "新しい JavaScript エラーが記録されたときにメールを送信します。デフォルトではオフです。",
+      errorEmailNotificationsSaveFailed: "メール設定を保存できませんでした。",
       about: "概要",
       aboutDescription:
         "Analytics はデータソースを接続し、カスタムダッシュボードを作成するツールです。Google Analytics、BigQuery、Stripe などを接続し、エージェントにダッシュボード作成を依頼できます。",
@@ -5361,6 +5389,10 @@ export const messagesByLocale = {
       openDashboardTemplates: "카탈로그 열기",
       languageTitle: "언어",
       languageLabel: "인터페이스 언어",
+      errorEmailNotifications: "새 오류 알림을 이메일로 받기",
+      errorEmailNotificationsDescription:
+        "새 JavaScript 오류가 캡처되면 이메일을 보냅니다. 기본값은 꺼져 있습니다.",
+      errorEmailNotificationsSaveFailed: "이메일 설정을 저장하지 못했습니다.",
       about: "정보",
       aboutDescription:
         "Analytics는 데이터 소스를 연결하고 사용자 지정 대시보드를 만드는 도구입니다. Google Analytics, BigQuery, Stripe 등을 연결한 뒤 에이전트에게 대시보드를 만들게 하세요.",
@@ -5584,6 +5616,11 @@ export const messagesByLocale = {
       openDashboardTemplates: "Abrir catálogo",
       languageTitle: "Idioma",
       languageLabel: "Idioma da interface",
+      errorEmailNotifications: "Receber novos alertas de erro por e-mail",
+      errorEmailNotificationsDescription:
+        "Envie um e-mail quando um novo erro de JavaScript for capturado. Desativado por padrão.",
+      errorEmailNotificationsSaveFailed:
+        "Não foi possível salvar a preferência de e-mail.",
       about: "Sobre",
       aboutDescription:
         "Analytics conecta fontes de dados e cria dashboards personalizados. Conecte Google Analytics, BigQuery, Stripe e outros, depois peça ao agente para criar dashboards.",
@@ -5802,6 +5839,10 @@ export const messagesByLocale = {
       openDashboardTemplates: "कैटलॉग खोलें",
       languageTitle: "भाषा",
       languageLabel: "इंटरफ़ेस भाषा",
+      errorEmailNotifications: "नए त्रुटि अलर्ट ईमेल से प्राप्त करें",
+      errorEmailNotificationsDescription:
+        "नया JavaScript त्रुटि कैप्चर होने पर ईमेल भेजें। डिफ़ॉल्ट रूप से बंद।",
+      errorEmailNotificationsSaveFailed: "ईमेल प्राथमिकता सहेजी नहीं जा सकी।",
       about: "परिचय",
       aboutDescription:
         "Analytics डेटा स्रोतों को जोड़ने और कस्टम डैशबोर्ड बनाने का टूल है। Google Analytics, BigQuery, Stripe आदि जोड़ें, फिर एजेंट से डैशबोर्ड बनवाएं।",
@@ -6019,6 +6060,11 @@ export const messagesByLocale = {
       openDashboardTemplates: "افتح الكتالوج",
       languageTitle: "اللغة",
       languageLabel: "لغة الواجهة",
+      errorEmailNotifications:
+        "تلقي تنبيهات الأخطاء الجديدة عبر البريد الإلكتروني",
+      errorEmailNotificationsDescription:
+        "إرسال بريد إلكتروني عند التقاط خطأ JavaScript جديد. معطّل افتراضيًا.",
+      errorEmailNotificationsSaveFailed: "تعذّر حفظ تفضيل البريد الإلكتروني.",
       about: "حول",
       aboutDescription:
         "Analytics أداة لربط مصادر البيانات وبناء لوحات معلومات مخصصة. اربط Google Analytics وBigQuery وStripe وغيرها، ثم اطلب من الوكيل إنشاء اللوحات.",

package/corpus/templates/analytics/app/pages/Settings.tsx

@@ -1,4 +1,8 @@
 import { ChangelogSettingsCard } from "@agent-native/core/client/changelog";
+import {
+  useActionMutation,
+  useActionQuery,
+} from "@agent-native/core/client/hooks";
 import { LanguagePicker, useT } from "@agent-native/core/client/i18n";
 import { TeamPage } from "@agent-native/core/client/org";
 import {
@@ -10,8 +14,9 @@ import {
   type SettingsTabItem,
 } from "@agent-native/core/client/settings";
 import { IconBell } from "@tabler/icons-react";
-import { useMemo } from "react";
+import { useEffect, useMemo, useState } from "react";
 import { Link } from "react-router";
+import { toast } from "sonner";
 
 import { Button } from "@/components/ui/button";
 import {
@@ -22,8 +27,13 @@ import {
   CardTitle,
 } from "@/components/ui/card";
 import { Label } from "@/components/ui/label";
+import { Switch } from "@/components/ui/switch";
 
 import changelog from "../../CHANGELOG.md?raw";
+import {
+  ANALYTICS_USER_PREFS_KEY,
+  type AnalyticsUserPrefs,
+} from "../../shared/analytics-user-prefs";
 import { useReplayStorageStatus } from "../hooks/use-replay-storage-status";
 import { ReplayStorageHint } from "./sessions/SessionsPage";
 import { AlertRulesSettingsCard } from "./settings/AlertRulesSettingsCard";
@@ -33,6 +43,47 @@ export default function Settings() {
   const t = useT();
   const agentSettingsTabs = useAgentSettingsTabs();
   const replayStorageStatus = useReplayStorageStatus();
+  const { data: analyticsPrefs, isLoading: analyticsPrefsLoading } =
+    useActionQuery<AnalyticsUserPrefs>("get-user-pref", {
+      key: ANALYTICS_USER_PREFS_KEY,
+    });
+  const saveAnalyticsPrefs = useActionMutation<
+    { success: boolean },
+    { key: string; value: Record<string, unknown> }
+  >("set-user-pref");
+  const [errorEmailEnabledOverride, setErrorEmailEnabledOverride] = useState<
+    boolean | null
+  >(null);
+
+  useEffect(() => {
+    if (analyticsPrefs) {
+      setErrorEmailEnabledOverride(
+        analyticsPrefs.errorEmailNotifications === true,
+      );
+    }
+  }, [analyticsPrefs]);
+
+  const errorEmailEnabled =
+    errorEmailEnabledOverride ??
+    analyticsPrefs?.errorEmailNotifications === true;
+
+  const saveErrorEmailPreference = (enabled: boolean) => {
+    const previous = errorEmailEnabled;
+    setErrorEmailEnabledOverride(enabled);
+    void saveAnalyticsPrefs
+      .mutateAsync({
+        key: ANALYTICS_USER_PREFS_KEY,
+        value: { errorEmailNotifications: enabled },
+      })
+      .catch((error) => {
+        setErrorEmailEnabledOverride(previous);
+        toast.error(
+          error instanceof Error
+            ? error.message
+            : t("settings.errorEmailNotificationsSaveFailed"),
+        );
+      });
+  };
 
   const extraTabs = useMemo<SettingsTabItem[]>(
     () => [
@@ -104,6 +155,21 @@ export default function Settings() {
                 </div>
               }
             />
+            <SettingsRow
+              id="error-email-notifications"
+              label={t("settings.errorEmailNotifications")}
+              description={t("settings.errorEmailNotificationsDescription")}
+              control={
+                <Switch
+                  aria-label={t("settings.errorEmailNotifications")}
+                  checked={errorEmailEnabled}
+                  disabled={
+                    analyticsPrefsLoading || saveAnalyticsPrefs.isPending
+                  }
+                  onCheckedChange={saveErrorEmailPreference}
+                />
+              }
+            />
           </SettingsGroup>
 
           {replayStorageStatus.data?.configured ? (

package/corpus/templates/analytics/app/pages/settings/settings-search.ts

@@ -51,6 +51,12 @@ export function buildAnalyticsGeneralSettingsSearchEntries(
       keywords: "language locale translation i18n",
       hash: "language",
     },
+    {
+      id: "analytics-error-email-notifications",
+      label: t("settings.errorEmailNotifications"),
+      keywords: "email notifications errors alerts javascript monitoring",
+      hash: "error-email-notifications",
+    },
     {
       id: "analytics-about",
       label: t("settings.about"),

package/corpus/templates/analytics/changelog/2026-08-03-dashboard-builds-now-finish-without-an-extra-confirmation.md

@@ -0,0 +1,6 @@
+---
+type: improved
+date: 2026-08-03
+---
+
+Dashboard requests now continue through data setup, saving, embedding, and navigation without stopping for an extra confirmation.

package/corpus/templates/analytics/changelog/2026-08-03-error-alert-emails-are-opt-in.md

@@ -0,0 +1,6 @@
+---
+type: changed
+date: 2026-08-03
+---
+
+New error alert emails are now off by default and can be enabled in Settings.

package/corpus/templates/analytics/server/lib/agent-chat-plan-mode.ts

@@ -13,20 +13,21 @@ export const INITIAL_TOOL_NAMES = [
   "list-session-recordings",
   "list-analyses",
   "get-analysis",
-  // Dashboard/extension INSPECTION stays on the initial surface so a
-  // template-clone request can resolve and inspect the source on the first
-  // turn. The MUTATING writers (update-dashboard, mutate-dashboard,
-  // create-extension, update-extension) are intentionally left off: the
-  // dashboard-construction final-response guard retries with
-  // `expandToolSurface: true` (see server/plugins/agent-chat.ts), which opens
-  // the full run registry exactly when a save is needed, and tool-search can
-  // surface them otherwise. This keeps the first-request surface under the
-  // 40-tool ceiling enforced by scripts/guard-agent-chat-context.ts.
+  // Keep the complete dashboard build path on the initial surface. An explicit
+  // build request should not stop at inspection or an empty extension shell
+  // while the agent lazily discovers the next mutating action.
   "get-sql-dashboard",
   "list-sql-dashboards",
   "list-dashboard-templates",
   "list-extensions",
   "get-extension",
+  "update-dashboard",
+  "mutate-dashboard",
+  "compose-dashboard",
+  "install-dashboard-template",
+  "create-extension",
+  "update-extension",
+  "extension-data-set",
   "generate-chart",
   "search-analytics-query-catalog",
   "query-agent-native-analytics",