@agent-native/core 0.145.8 → 0.146.1

package/corpus/templates/dispatch/.agents/skills/turn-into-app/SKILL.md

@@ -4,7 +4,7 @@ description: >-
   Turn visible project context, a proven thread, skill, or workflow into a
   runnable Agent-Native app with simple buttons, visible agent steps, preview,
   and deployment handoff. Use when a user invokes `/turn-into-app` or asks to
-  make a workflow into an app.
+  make a workflow into an app, including from Claude or ChatGPT on the web.
 user-invocable: true
 scope: both
 metadata:
@@ -13,9 +13,36 @@ metadata:
 
 # Turn Into App
 
+## Host execution boundary
+
+The build location depends on the host. Treat this as a hard routing rule:
+
+- In Claude Web, ChatGPT Web, or a Claude/ChatGPT Project on the web, you are
+  the source analyst and handoff orchestrator. You must not build the app in
+  the host sandbox. Do not run `npm`, `pnpm`, `npx`, `agent-native create`, or
+  `add-app`; do not edit files, create artifacts, or start a local dev server.
+  After writing the bounded source brief, call the connected Dispatch action
+  `start-workspace-app-creation`. Pass the brief and repeatable workflow in
+  `prompt`, plus the inferred `appId`, `description`, `template`, and selected
+  `resourceIds` when available. This is the Builder handoff.
+- Do not substitute the generic `create_workspace_app` MCP tool in an online
+  host. That tool is a local workspace scaffolder, not the Builder handoff.
+- Connect the Agent-Native Dispatch MCP connector only. Dispatch uses the
+  authenticated Builder Projects API to reuse or provision the workspace
+  project before starting the Builder Cloud Agent; a separate Builder CMS MCP
+  connection is not required for this workflow.
+- If `start-workspace-app-creation` is not available or Dispatch is not
+  authenticated, stop with the connector setup needed. Do not fall back to a
+  local build or claim that the app exists.
+- In Claude Code, Codex Code, or another local code-agent runtime with a target
+  workspace, follow the local implementation steps below. The local agent may
+  scaffold, edit, run, and verify the app there.
+
 ## Default behavior
 
-This is an end-to-end build skill, not a request for an app proposal.
+For a local code-agent runtime this is an end-to-end build skill, not a request
+for an app proposal. For an online host, the end-to-end result is a verified
+Builder handoff and the resulting workspace app, not code written in the host.
 
 - With no argument, choose the source in this order: visible project context,
   then the current thread. A fresh Claude or ChatGPT Project is a valid source
@@ -47,9 +74,12 @@ private web access, invent an importer, add fake OAuth, or scrape a logged-in
 page. If the needed context is not visible, ask for an export, transcript, or
 attachment and treat that artifact as imported source material.
 
-Deliver a fresh app in a new directory, implement the repeatable workflow with
-buttons and agent handoffs, start its dev server, verify the main path, and
-continue through build/deployment handoff. Do not stop at a plan.
+For local code-agent runtimes, deliver a fresh app in a new directory,
+implement the repeatable workflow with buttons and agent handoffs, start its dev
+server, verify the main path, and continue through build/deployment handoff. Do
+not stop at a plan. For online hosts, call Dispatch first after the source brief,
+then report the returned Builder branch/path and verify the workspace app through
+Dispatch when it becomes available.
 
 ## Fresh project context mode
 
@@ -190,12 +220,12 @@ Do not use `create` for an existing workspace; it scaffolds a new standalone
 workspace rather than adding an app to the current one.
 
 When the source came from a fresh external Project and the target is a Builder
-workspace, prefer the available Dispatch MCP workspace handoff after writing
-the source brief. Call `start-workspace-app-creation` with a concise prompt,
-the inferred app id, and the brief's repeatable workflow. Pass selected
-workspace resource IDs when they are available, rather than pasting entire
-knowledge files. Continue from the returned workspace path or Builder handoff
-and report the actual verification result.
+workspace, the online host path is mandatory: after writing the source brief,
+call `start-workspace-app-creation` with a concise prompt, the inferred app id,
+and the brief's repeatable workflow. Pass selected workspace resource IDs when
+they are available, rather than pasting entire knowledge files. Continue from
+the returned workspace path or Builder branch handoff and report the actual
+verification result. Do not run the local scaffold commands in this host mode.
 
 Use a first-party template only when it materially fits the workflow. Keep the
 new app independent from the source thread's working tree unless the user

package/corpus/templates/dispatch/.agents/skills/turn-into-app/references/fresh-project.md

@@ -50,14 +50,33 @@ skills. For a ChatGPT Project or any host that exposes only the MCP connector,
 also place the skill instructions or this reference in the Project
 instructions/knowledge files. Then add and authenticate the Dispatch MCP
 connector from the packaged `adapters/chatgpt-mcp/connector.json`.
+This single connector is enough: Dispatch provisions or reuses the Builder
+project through the Builder Projects API, so do not ask the user to add a
+separate Builder CMS MCP for app creation.
+Builder's separate Fusion MCP is available as a custom remote connector at
+`https://mcp.builder.io/mcp/fusion` for direct Builder work. It can run an
+existing project, but it cannot provision the repo-backed Agent-Native
+workspace project, so it is optional and not a replacement for the Dispatch
+handoff in this skill. Do not add Builder's CMS MCP for this workflow.
+
+For Claude Web, ChatGPT Web, and their web Projects, the host is an
+orchestrator, not the build environment. After forming the source brief, call
+the Dispatch `start-workspace-app-creation` action so Builder creates the app
+in the connected workspace. Do not use the host's code interpreter, shell,
+artifact editor, `create_workspace_app`, or local `pnpm`/`npm` commands. If the
+handoff action is unavailable, stop and ask the user to authenticate Dispatch;
+do not fall back to a local sandbox build. Claude Code and Codex Code are the
+local-agent exception and may follow the implementation steps in the main
+skill.
 
 In a new Project chat, say:
 
 ```text
 Turn this project into an app. Use the visible Project instructions,
 knowledge files, and any selected successful runs as the source. Create the
-app in the connected Agent-Native workspace, keep the source brief bounded,
-and report the real path and verification result.
+app in the connected Agent-Native workspace through Dispatch and Builder, keep
+the source brief bounded, and report the real Builder branch/path and
+verification result. Do not build it in this chat's sandbox.
 ```
 
 If the host does not show the Project instructions or files to the model, stop

package/corpus/templates/dispatch/changelog/2026-08-07-language-selection-now-lives-in-settings-instead-of-the-disp.md

@@ -0,0 +1,6 @@
+---
+type: changed
+date: 2026-08-07
+---
+
+Language selection now lives in Settings instead of the Dispatch header.

package/corpus/templates/factory/.agents/skills/turn-into-app/SKILL.md

@@ -4,7 +4,7 @@ description: >-
   Turn visible project context, a proven thread, skill, or workflow into a
   runnable Agent-Native app with simple buttons, visible agent steps, preview,
   and deployment handoff. Use when a user invokes `/turn-into-app` or asks to
-  make a workflow into an app.
+  make a workflow into an app, including from Claude or ChatGPT on the web.
 user-invocable: true
 scope: both
 metadata:
@@ -13,9 +13,36 @@ metadata:
 
 # Turn Into App
 
+## Host execution boundary
+
+The build location depends on the host. Treat this as a hard routing rule:
+
+- In Claude Web, ChatGPT Web, or a Claude/ChatGPT Project on the web, you are
+  the source analyst and handoff orchestrator. You must not build the app in
+  the host sandbox. Do not run `npm`, `pnpm`, `npx`, `agent-native create`, or
+  `add-app`; do not edit files, create artifacts, or start a local dev server.
+  After writing the bounded source brief, call the connected Dispatch action
+  `start-workspace-app-creation`. Pass the brief and repeatable workflow in
+  `prompt`, plus the inferred `appId`, `description`, `template`, and selected
+  `resourceIds` when available. This is the Builder handoff.
+- Do not substitute the generic `create_workspace_app` MCP tool in an online
+  host. That tool is a local workspace scaffolder, not the Builder handoff.
+- Connect the Agent-Native Dispatch MCP connector only. Dispatch uses the
+  authenticated Builder Projects API to reuse or provision the workspace
+  project before starting the Builder Cloud Agent; a separate Builder CMS MCP
+  connection is not required for this workflow.
+- If `start-workspace-app-creation` is not available or Dispatch is not
+  authenticated, stop with the connector setup needed. Do not fall back to a
+  local build or claim that the app exists.
+- In Claude Code, Codex Code, or another local code-agent runtime with a target
+  workspace, follow the local implementation steps below. The local agent may
+  scaffold, edit, run, and verify the app there.
+
 ## Default behavior
 
-This is an end-to-end build skill, not a request for an app proposal.
+For a local code-agent runtime this is an end-to-end build skill, not a request
+for an app proposal. For an online host, the end-to-end result is a verified
+Builder handoff and the resulting workspace app, not code written in the host.
 
 - With no argument, choose the source in this order: visible project context,
   then the current thread. A fresh Claude or ChatGPT Project is a valid source
@@ -47,9 +74,12 @@ private web access, invent an importer, add fake OAuth, or scrape a logged-in
 page. If the needed context is not visible, ask for an export, transcript, or
 attachment and treat that artifact as imported source material.
 
-Deliver a fresh app in a new directory, implement the repeatable workflow with
-buttons and agent handoffs, start its dev server, verify the main path, and
-continue through build/deployment handoff. Do not stop at a plan.
+For local code-agent runtimes, deliver a fresh app in a new directory,
+implement the repeatable workflow with buttons and agent handoffs, start its dev
+server, verify the main path, and continue through build/deployment handoff. Do
+not stop at a plan. For online hosts, call Dispatch first after the source brief,
+then report the returned Builder branch/path and verify the workspace app through
+Dispatch when it becomes available.
 
 ## Fresh project context mode
 
@@ -190,12 +220,12 @@ Do not use `create` for an existing workspace; it scaffolds a new standalone
 workspace rather than adding an app to the current one.
 
 When the source came from a fresh external Project and the target is a Builder
-workspace, prefer the available Dispatch MCP workspace handoff after writing
-the source brief. Call `start-workspace-app-creation` with a concise prompt,
-the inferred app id, and the brief's repeatable workflow. Pass selected
-workspace resource IDs when they are available, rather than pasting entire
-knowledge files. Continue from the returned workspace path or Builder handoff
-and report the actual verification result.
+workspace, the online host path is mandatory: after writing the source brief,
+call `start-workspace-app-creation` with a concise prompt, the inferred app id,
+and the brief's repeatable workflow. Pass selected workspace resource IDs when
+they are available, rather than pasting entire knowledge files. Continue from
+the returned workspace path or Builder branch handoff and report the actual
+verification result. Do not run the local scaffold commands in this host mode.
 
 Use a first-party template only when it materially fits the workflow. Keep the
 new app independent from the source thread's working tree unless the user

package/corpus/templates/factory/.agents/skills/turn-into-app/references/fresh-project.md

@@ -50,14 +50,33 @@ skills. For a ChatGPT Project or any host that exposes only the MCP connector,
 also place the skill instructions or this reference in the Project
 instructions/knowledge files. Then add and authenticate the Dispatch MCP
 connector from the packaged `adapters/chatgpt-mcp/connector.json`.
+This single connector is enough: Dispatch provisions or reuses the Builder
+project through the Builder Projects API, so do not ask the user to add a
+separate Builder CMS MCP for app creation.
+Builder's separate Fusion MCP is available as a custom remote connector at
+`https://mcp.builder.io/mcp/fusion` for direct Builder work. It can run an
+existing project, but it cannot provision the repo-backed Agent-Native
+workspace project, so it is optional and not a replacement for the Dispatch
+handoff in this skill. Do not add Builder's CMS MCP for this workflow.
+
+For Claude Web, ChatGPT Web, and their web Projects, the host is an
+orchestrator, not the build environment. After forming the source brief, call
+the Dispatch `start-workspace-app-creation` action so Builder creates the app
+in the connected workspace. Do not use the host's code interpreter, shell,
+artifact editor, `create_workspace_app`, or local `pnpm`/`npm` commands. If the
+handoff action is unavailable, stop and ask the user to authenticate Dispatch;
+do not fall back to a local sandbox build. Claude Code and Codex Code are the
+local-agent exception and may follow the implementation steps in the main
+skill.
 
 In a new Project chat, say:
 
 ```text
 Turn this project into an app. Use the visible Project instructions,
 knowledge files, and any selected successful runs as the source. Create the
-app in the connected Agent-Native workspace, keep the source brief bounded,
-and report the real path and verification result.
+app in the connected Agent-Native workspace through Dispatch and Builder, keep
+the source brief bounded, and report the real Builder branch/path and
+verification result. Do not build it in this chat's sandbox.
 ```
 
 If the host does not show the Project instructions or files to the model, stop

package/corpus/templates/factory/server/plugins/factory-scheduler-job.ts

@@ -169,9 +169,10 @@ const AUTOMATION_SEEDS: AutomationSeed[] = [
 # Factory Slack feedback triage
 
 Read the Factory configuration. When Slack polling is enabled and a channel is
-configured, call poll-slack-channel first. Then list a bounded page of
-received Slack items and call get-slack-feedback-context for each item before
-classifying it.
+configured, call poll-slack-channel first. Then list at most 2 received Slack
+items by passing limit 2, process them sequentially, and call
+get-slack-feedback-context for each item before classifying it. Never use the
+action's default page size.
 
 Start work only for a clear bug: a concrete broken behavior, reproducible
 failure, error, regression, stuck run, incorrect result, or a report with a
@@ -203,9 +204,9 @@ PR, merge, or fix unless an action returned that state.
 # Factory Sentry error triage
 
 Read the Factory configuration. When Sentry polling is enabled and a Sentry
-organization is configured, call poll-sentry-errors. List a bounded page of
-received Sentry items and inspect the title, culprit, level, event count, and
-errorReport metadata.
+organization is configured, call poll-sentry-errors. List at most 3 received
+Sentry items by passing limit 3 and inspect the title, culprit, level, event
+count, and errorReport metadata. Never use the action's default page size.
 
 Only classify a concrete unresolved error as a clear bug when the Sentry
 evidence is sufficient to investigate. Do not dispatch on noise, expected
@@ -226,7 +227,8 @@ run callback or PR observation confirms it.
 
 Read the Factory configuration. When GitHub source polling is enabled and a
 repository is configured, call poll-github-sources with includeIssues true and
-includePullRequests false. List a bounded page of received github_issue items.
+includePullRequests false. List at most 3 received github_issue items by
+passing limit 3. Never use the action's default page size.
 
 Treat an issue as a clear bug only when it has a concrete error report,
 reproduction, incorrect behavior, regression, or specific failing path. Do
@@ -247,7 +249,8 @@ confirmation.
 
 Read the Factory configuration. When GitHub polling is enabled and a repository
 is configured, call poll-github-sources with includeIssues false and
-includePullRequests true. List a bounded page of github pull-request items.
+includePullRequests true. List at most 3 github pull-request items by passing
+limit 3. Never use the action's default page size.
 
 For each open agent-native PR, inspect the item and classify whether it is a
 clear bug fix or has product or UX implications. Call
@@ -275,7 +278,8 @@ governance action confirms it.
 
 Read the Factory configuration. When GitHub polling is enabled and a repository
 is configured, call poll-github-sources with includeIssues false and
-includePullRequests true. List a bounded page of open github pull-request items.
+includePullRequests true. List at most 3 open github pull-request items by
+passing limit 3. Never use the action's default page size.
 
 For each item, call babysit-agent-native-pull-request. That action fetches fresh
 GitHub and ai-services evidence and is the only place allowed to decide whether
@@ -312,6 +316,22 @@ function defaultGithubPollingEnabled(): 0 | 1 {
     : 0;
 }
 
+function automationPromptGuard(name: string): string | undefined {
+  switch (name) {
+    case "factory-slack-feedback":
+      return "Runtime safety bound: pass limit 2 to list-triage-items, process at most two Slack items sequentially, and never use the default page size.";
+    case "factory-sentry-errors":
+      return "Runtime safety bound: pass limit 3 to list-triage-items and process at most three Sentry items.";
+    case "factory-github-issues":
+      return "Runtime safety bound: pass limit 3 to list-triage-items and process at most three GitHub issue items.";
+    case "factory-pr-governance":
+    case "factory-pr-babysit":
+      return "Runtime safety bound: pass limit 3 to list-triage-items and process at most three pull-request items.";
+    default:
+      return undefined;
+  }
+}
+
 function setFrontmatterField(
   content: string,
   key: string,
@@ -388,6 +408,10 @@ async function ensureOrganizationAutomations(
       repaired = setFrontmatterField(repaired, "orgId", orgId);
       repaired = setFrontmatterField(repaired, "createdBy", ownerEmail);
       repaired = setFrontmatterField(repaired, "runAs", "creator");
+      const promptGuard = automationPromptGuard(seed.name);
+      if (promptGuard && !repaired.includes(promptGuard)) {
+        repaired = `${repaired.trimEnd()}\n\n${promptGuard}\n`;
+      }
       if (repaired === existing.content) return;
 
       const updated = await resourcePutIfCurrent({
@@ -415,11 +439,16 @@ async function ensureDefaultTriageConfig(
   const db = getDb();
   const existing = (
     await db
-      .select({ id: triageConfig.id })
+      .select({
+        id: triageConfig.id,
+      })
       .from(triageConfig)
       .where(and(eq(triageConfig.id, orgId), eq(triageConfig.orgId, orgId)))
       .limit(1)
   )[0];
+  const repository = defaultRepository();
+  // Existing rows are operator-owned. An empty repository plus disabled
+  // polling can be an intentional choice, so do not infer bootstrap state.
   if (existing) return;
   const now = new Date().toISOString();
   await db.insert(triageConfig).values({
@@ -432,7 +461,7 @@ async function ensureDefaultTriageConfig(
     sentryPollingEnabled: 0,
     lastSlackTs: null,
     slackHistoryCursor: null,
-    repository: defaultRepository(),
+    repository,
     sentryOrgSlug: null,
     sentryProjectSlug: null,
     sentryEnvironment: null,
@@ -445,17 +474,25 @@ async function ensureDefaultTriageConfig(
 }
 
 async function ensureSchedulerJobs(): Promise<void> {
-  const ownerEmail = workspaceOwnerEmail();
-  if (!ownerEmail) {
-    throw new Error(
-      "WORKSPACE_OWNER_EMAIL is required to seed Factory automations",
-    );
-  }
-  const orgId = await resolveOrgIdForEmail(ownerEmail);
-  if (!orgId) {
-    throw new Error(
-      "The Factory deployment owner must belong to an organization before automations can be seeded",
-    );
+  let ownerEmail = workspaceOwnerEmail();
+  let orgId = ownerEmail ? await resolveOrgIdForEmail(ownerEmail) : undefined;
+  if (!ownerEmail || !orgId) {
+    const existingConfigs = await getDb()
+      .select({
+        id: triageConfig.id,
+        ownerEmail: triageConfig.ownerEmail,
+        orgId: triageConfig.orgId,
+      })
+      .from(triageConfig)
+      .limit(2);
+    if (existingConfigs.length !== 1) {
+      throw new Error(
+        "WORKSPACE_OWNER_EMAIL is required to seed Factory automations when the Factory organization is not uniquely configured",
+      );
+    }
+    const existingConfig = existingConfigs[0];
+    ownerEmail = existingConfig.ownerEmail.trim().toLowerCase();
+    orgId = existingConfig.orgId?.trim() || existingConfig.id;
   }
   await ensureDefaultTriageConfig(ownerEmail, orgId);
   await ensureOrganizationAutomations(ownerEmail, orgId);

package/corpus/templates/forms/.agents/skills/turn-into-app/SKILL.md

@@ -4,7 +4,7 @@ description: >-
   Turn visible project context, a proven thread, skill, or workflow into a
   runnable Agent-Native app with simple buttons, visible agent steps, preview,
   and deployment handoff. Use when a user invokes `/turn-into-app` or asks to
-  make a workflow into an app.
+  make a workflow into an app, including from Claude or ChatGPT on the web.
 user-invocable: true
 scope: both
 metadata:
@@ -13,9 +13,36 @@ metadata:
 
 # Turn Into App
 
+## Host execution boundary
+
+The build location depends on the host. Treat this as a hard routing rule:
+
+- In Claude Web, ChatGPT Web, or a Claude/ChatGPT Project on the web, you are
+  the source analyst and handoff orchestrator. You must not build the app in
+  the host sandbox. Do not run `npm`, `pnpm`, `npx`, `agent-native create`, or
+  `add-app`; do not edit files, create artifacts, or start a local dev server.
+  After writing the bounded source brief, call the connected Dispatch action
+  `start-workspace-app-creation`. Pass the brief and repeatable workflow in
+  `prompt`, plus the inferred `appId`, `description`, `template`, and selected
+  `resourceIds` when available. This is the Builder handoff.
+- Do not substitute the generic `create_workspace_app` MCP tool in an online
+  host. That tool is a local workspace scaffolder, not the Builder handoff.
+- Connect the Agent-Native Dispatch MCP connector only. Dispatch uses the
+  authenticated Builder Projects API to reuse or provision the workspace
+  project before starting the Builder Cloud Agent; a separate Builder CMS MCP
+  connection is not required for this workflow.
+- If `start-workspace-app-creation` is not available or Dispatch is not
+  authenticated, stop with the connector setup needed. Do not fall back to a
+  local build or claim that the app exists.
+- In Claude Code, Codex Code, or another local code-agent runtime with a target
+  workspace, follow the local implementation steps below. The local agent may
+  scaffold, edit, run, and verify the app there.
+
 ## Default behavior
 
-This is an end-to-end build skill, not a request for an app proposal.
+For a local code-agent runtime this is an end-to-end build skill, not a request
+for an app proposal. For an online host, the end-to-end result is a verified
+Builder handoff and the resulting workspace app, not code written in the host.
 
 - With no argument, choose the source in this order: visible project context,
   then the current thread. A fresh Claude or ChatGPT Project is a valid source
@@ -47,9 +74,12 @@ private web access, invent an importer, add fake OAuth, or scrape a logged-in
 page. If the needed context is not visible, ask for an export, transcript, or
 attachment and treat that artifact as imported source material.
 
-Deliver a fresh app in a new directory, implement the repeatable workflow with
-buttons and agent handoffs, start its dev server, verify the main path, and
-continue through build/deployment handoff. Do not stop at a plan.
+For local code-agent runtimes, deliver a fresh app in a new directory,
+implement the repeatable workflow with buttons and agent handoffs, start its dev
+server, verify the main path, and continue through build/deployment handoff. Do
+not stop at a plan. For online hosts, call Dispatch first after the source brief,
+then report the returned Builder branch/path and verify the workspace app through
+Dispatch when it becomes available.
 
 ## Fresh project context mode
 
@@ -190,12 +220,12 @@ Do not use `create` for an existing workspace; it scaffolds a new standalone
 workspace rather than adding an app to the current one.
 
 When the source came from a fresh external Project and the target is a Builder
-workspace, prefer the available Dispatch MCP workspace handoff after writing
-the source brief. Call `start-workspace-app-creation` with a concise prompt,
-the inferred app id, and the brief's repeatable workflow. Pass selected
-workspace resource IDs when they are available, rather than pasting entire
-knowledge files. Continue from the returned workspace path or Builder handoff
-and report the actual verification result.
+workspace, the online host path is mandatory: after writing the source brief,
+call `start-workspace-app-creation` with a concise prompt, the inferred app id,
+and the brief's repeatable workflow. Pass selected workspace resource IDs when
+they are available, rather than pasting entire knowledge files. Continue from
+the returned workspace path or Builder branch handoff and report the actual
+verification result. Do not run the local scaffold commands in this host mode.
 
 Use a first-party template only when it materially fits the workflow. Keep the
 new app independent from the source thread's working tree unless the user

package/corpus/templates/forms/.agents/skills/turn-into-app/references/fresh-project.md

@@ -50,14 +50,33 @@ skills. For a ChatGPT Project or any host that exposes only the MCP connector,
 also place the skill instructions or this reference in the Project
 instructions/knowledge files. Then add and authenticate the Dispatch MCP
 connector from the packaged `adapters/chatgpt-mcp/connector.json`.
+This single connector is enough: Dispatch provisions or reuses the Builder
+project through the Builder Projects API, so do not ask the user to add a
+separate Builder CMS MCP for app creation.
+Builder's separate Fusion MCP is available as a custom remote connector at
+`https://mcp.builder.io/mcp/fusion` for direct Builder work. It can run an
+existing project, but it cannot provision the repo-backed Agent-Native
+workspace project, so it is optional and not a replacement for the Dispatch
+handoff in this skill. Do not add Builder's CMS MCP for this workflow.
+
+For Claude Web, ChatGPT Web, and their web Projects, the host is an
+orchestrator, not the build environment. After forming the source brief, call
+the Dispatch `start-workspace-app-creation` action so Builder creates the app
+in the connected workspace. Do not use the host's code interpreter, shell,
+artifact editor, `create_workspace_app`, or local `pnpm`/`npm` commands. If the
+handoff action is unavailable, stop and ask the user to authenticate Dispatch;
+do not fall back to a local sandbox build. Claude Code and Codex Code are the
+local-agent exception and may follow the implementation steps in the main
+skill.
 
 In a new Project chat, say:
 
 ```text
 Turn this project into an app. Use the visible Project instructions,
 knowledge files, and any selected successful runs as the source. Create the
-app in the connected Agent-Native workspace, keep the source brief bounded,
-and report the real path and verification result.
+app in the connected Agent-Native workspace through Dispatch and Builder, keep
+the source brief bounded, and report the real Builder branch/path and
+verification result. Do not build it in this chat's sandbox.
 ```
 
 If the host does not show the Project instructions or files to the model, stop

package/corpus/templates/macros/.agents/skills/turn-into-app/SKILL.md

@@ -4,7 +4,7 @@ description: >-
   Turn visible project context, a proven thread, skill, or workflow into a
   runnable Agent-Native app with simple buttons, visible agent steps, preview,
   and deployment handoff. Use when a user invokes `/turn-into-app` or asks to
-  make a workflow into an app.
+  make a workflow into an app, including from Claude or ChatGPT on the web.
 user-invocable: true
 scope: both
 metadata:
@@ -13,9 +13,36 @@ metadata:
 
 # Turn Into App
 
+## Host execution boundary
+
+The build location depends on the host. Treat this as a hard routing rule:
+
+- In Claude Web, ChatGPT Web, or a Claude/ChatGPT Project on the web, you are
+  the source analyst and handoff orchestrator. You must not build the app in
+  the host sandbox. Do not run `npm`, `pnpm`, `npx`, `agent-native create`, or
+  `add-app`; do not edit files, create artifacts, or start a local dev server.
+  After writing the bounded source brief, call the connected Dispatch action
+  `start-workspace-app-creation`. Pass the brief and repeatable workflow in
+  `prompt`, plus the inferred `appId`, `description`, `template`, and selected
+  `resourceIds` when available. This is the Builder handoff.
+- Do not substitute the generic `create_workspace_app` MCP tool in an online
+  host. That tool is a local workspace scaffolder, not the Builder handoff.
+- Connect the Agent-Native Dispatch MCP connector only. Dispatch uses the
+  authenticated Builder Projects API to reuse or provision the workspace
+  project before starting the Builder Cloud Agent; a separate Builder CMS MCP
+  connection is not required for this workflow.
+- If `start-workspace-app-creation` is not available or Dispatch is not
+  authenticated, stop with the connector setup needed. Do not fall back to a
+  local build or claim that the app exists.
+- In Claude Code, Codex Code, or another local code-agent runtime with a target
+  workspace, follow the local implementation steps below. The local agent may
+  scaffold, edit, run, and verify the app there.
+
 ## Default behavior
 
-This is an end-to-end build skill, not a request for an app proposal.
+For a local code-agent runtime this is an end-to-end build skill, not a request
+for an app proposal. For an online host, the end-to-end result is a verified
+Builder handoff and the resulting workspace app, not code written in the host.
 
 - With no argument, choose the source in this order: visible project context,
   then the current thread. A fresh Claude or ChatGPT Project is a valid source
@@ -47,9 +74,12 @@ private web access, invent an importer, add fake OAuth, or scrape a logged-in
 page. If the needed context is not visible, ask for an export, transcript, or
 attachment and treat that artifact as imported source material.
 
-Deliver a fresh app in a new directory, implement the repeatable workflow with
-buttons and agent handoffs, start its dev server, verify the main path, and
-continue through build/deployment handoff. Do not stop at a plan.
+For local code-agent runtimes, deliver a fresh app in a new directory,
+implement the repeatable workflow with buttons and agent handoffs, start its dev
+server, verify the main path, and continue through build/deployment handoff. Do
+not stop at a plan. For online hosts, call Dispatch first after the source brief,
+then report the returned Builder branch/path and verify the workspace app through
+Dispatch when it becomes available.
 
 ## Fresh project context mode
 
@@ -190,12 +220,12 @@ Do not use `create` for an existing workspace; it scaffolds a new standalone
 workspace rather than adding an app to the current one.
 
 When the source came from a fresh external Project and the target is a Builder
-workspace, prefer the available Dispatch MCP workspace handoff after writing
-the source brief. Call `start-workspace-app-creation` with a concise prompt,
-the inferred app id, and the brief's repeatable workflow. Pass selected
-workspace resource IDs when they are available, rather than pasting entire
-knowledge files. Continue from the returned workspace path or Builder handoff
-and report the actual verification result.
+workspace, the online host path is mandatory: after writing the source brief,
+call `start-workspace-app-creation` with a concise prompt, the inferred app id,
+and the brief's repeatable workflow. Pass selected workspace resource IDs when
+they are available, rather than pasting entire knowledge files. Continue from
+the returned workspace path or Builder branch handoff and report the actual
+verification result. Do not run the local scaffold commands in this host mode.
 
 Use a first-party template only when it materially fits the workflow. Keep the
 new app independent from the source thread's working tree unless the user

package/corpus/templates/macros/.agents/skills/turn-into-app/references/fresh-project.md

@@ -50,14 +50,33 @@ skills. For a ChatGPT Project or any host that exposes only the MCP connector,
 also place the skill instructions or this reference in the Project
 instructions/knowledge files. Then add and authenticate the Dispatch MCP
 connector from the packaged `adapters/chatgpt-mcp/connector.json`.
+This single connector is enough: Dispatch provisions or reuses the Builder
+project through the Builder Projects API, so do not ask the user to add a
+separate Builder CMS MCP for app creation.
+Builder's separate Fusion MCP is available as a custom remote connector at
+`https://mcp.builder.io/mcp/fusion` for direct Builder work. It can run an
+existing project, but it cannot provision the repo-backed Agent-Native
+workspace project, so it is optional and not a replacement for the Dispatch
+handoff in this skill. Do not add Builder's CMS MCP for this workflow.
+
+For Claude Web, ChatGPT Web, and their web Projects, the host is an
+orchestrator, not the build environment. After forming the source brief, call
+the Dispatch `start-workspace-app-creation` action so Builder creates the app
+in the connected workspace. Do not use the host's code interpreter, shell,
+artifact editor, `create_workspace_app`, or local `pnpm`/`npm` commands. If the
+handoff action is unavailable, stop and ask the user to authenticate Dispatch;
+do not fall back to a local sandbox build. Claude Code and Codex Code are the
+local-agent exception and may follow the implementation steps in the main
+skill.
 
 In a new Project chat, say:
 
 ```text
 Turn this project into an app. Use the visible Project instructions,
 knowledge files, and any selected successful runs as the source. Create the
-app in the connected Agent-Native workspace, keep the source brief bounded,
-and report the real path and verification result.
+app in the connected Agent-Native workspace through Dispatch and Builder, keep
+the source brief bounded, and report the real Builder branch/path and
+verification result. Do not build it in this chat's sandbox.
 ```
 
 If the host does not show the Project instructions or files to the model, stop