@salesforce/afv-skills 1.65.0 → 1.65.1

package/skills/automotive-cloud-sales-concierge-foundation-coordinate/SKILL.md

@@ -1,159 +0,0 @@
----
-name: automotive-cloud-sales-concierge-foundation-coordinate
-description: "Set up the Automotive Sales Concierge Agent end to end in a new org. Configure Automotive Sales Concierge from scratch. Stand up the Automotive Sales Concierge foundation completely. This is the single top-level orchestrator that runs all setup Acts in dependency order — Act 0 (enable Automotive Cloud org preferences + assign permission sets), Act 1 (create and activate the Automotive Sales Concierge NGA agent), and Act 2 (ground the agent's search infrastructure: DPE, CBSF, FlexiPages) — headlessly and API-first. TRIGGER when the user wants to \"set up the Automotive Sales Concierge\", \"setup Automotive Sales Concierge end to end\", \"help me setup Automotive Sales Concierge Agent\", \"stand up the Automotive Sales Concierge agent in this org\", \"configure the Automotive Sales Concierge foundation\", or otherwise asks for the whole multi-act Automotive Sales Concierge setup rather than a single narrow step. DO NOT TRIGGER when the user asks for exactly one specific toggle, one specific act, or for non-setup work (creating data, running DPE jobs, messaging channels)."
-metadata:
-  version: "1.0"
----
-
-# Automotive Sales Concierge Foundation — Orchestrator
-
-Stand up the **setup layer** of the Automotive Sales Concierge steel thread by running its
-enablement **Acts in a fixed, dependency-respecting order**, headlessly and API-first, against a
-live org. This skill is the single entry point a user (or a higher-level agent) invokes to get the
-whole foundation in place; it delegates the actual mutations to focused sub-skills and rolls their
-results into one report.
-
-**Scope: foundation setup + agent creation + search infrastructure grounding.** The Acts this
-orchestrator runs cover setup activities (enabling org preferences / feature toggles, assigning
-permission sets), agent lifecycle (creating the agent from the NGA template, verifying its structure,
-activating it), and grounding the agent's search infrastructure (Act 2): cloning + activating DPE
-definitions, configuring Criteria-Based Search & Filter (CBSF) for 3 searchable objects, and
-deploying CBSF FlexiPages — delegated sequentially to `mfg-spartans/dpe-clone-activate`,
-`mfg-knights/CriteriaBasedSearchAndFilter`, and `mfg-knights/activate-flexipage-via-action-override`.
-Creating business data, **running** the Data Pipelines (DPE) jobs, and configuring messaging channels
-are **not** part of this orchestrator; they are separate downstream skills invoked after the agent is
-live.
-
----
-
-## What this orchestrator runs today
-
-| Order | Act | Sub-skill | What it does |
-|---|---|---|---|
-| 1 | **Act 0 — Enable the Foundation** | `enable-automotive-sales-concierge-foundation` | Enables the Agentforce platform prerequisites (Einstein GPT Platform, Agent Platform, Einstein Copilot) then turns on the Automotive Cloud org preferences (Automotive Cloud, Scheduler, Agents, Appraisals, CBSF, Appraisal Mgmt, Partner Lead Mgmt, Industries Field Service, Quotes) via the Metadata API at `/services/data/v<apiVersion>/headless/metadata`; hands off the legally-gated Automotive Generative AI preference to the user for UI enablement (HARD STOP — waits for user confirmation); and assigns the foundation permission sets. |
-| 2 | **Act 1 — Create & Activate the Agent** | `create-automotive-sales-concierge-agent` | Creates the Automotive Sales Concierge agent from the pre-installed NGA template (`AutoGenAi__AutomotiveSalesConcierge`) via NextGen Authoring (NGA) bundle creation — resolving the template's agentScript, creating an NGA bundle, publishing, and activating. Verifies the subagent structure and all 13 expected actions are present, and activates the agent version so it is live for sales reps. |
-| 3 | **Act 2 — Ground the Agent's Search Infrastructure** | `ground-agent-data-pipelines` | Grounds the Sales Concierge's search infrastructure via 3 delegates in sequence: (1) `mfg-spartans/dpe-clone-activate` clones + activates 3 DPE definitions; (2) `mfg-knights/CriteriaBasedSearchAndFilter` creates SearchCriteriaConfiguration + mappings + SearchableObjDataSyncInfo for 3 searchable objects (VehDefSearchableField, DealerProdtSearchableField, DealerVehDefSearchableField); (3) `mfg-knights/activate-flexipage-via-action-override` deploys RecordPage FlexiPages with the CBSF component and activates them inside the Automotive app. A thin delegator: Act 2 owns only the domain data (which templates, which objects, which fields) and sequences the delegates with data threading (DPE 9N1 Ids → CBSF DataSyncJobId; CBSF DeveloperName → FlexiPage searchConfigs). |
-
-Today the orchestrator runs **three Acts (Act 0 → Act 1 → Act 2)**. Act 0 is the foundation every later
-Act depends on; Act 1 creates and activates the agent that Act 0 made possible; Act 2 grounds that
-agent's search infrastructure — cloning DPE definitions, configuring CBSF search for 3 searchable
-objects, and deploying the CBSF FlexiPages inside the Automotive app. Shipping the orchestrator with
-this ordered-run contract means future Acts slot in without re-plumbing.
-
-## What this orchestrator is designed to run next
-
-The orchestrator is built to grow. Future setup Acts become **additional sub-skills** under this
-workflow, and this skill runs them **in order, after their dependencies succeed**.
-
-CBSF enablement (previously planned as a separate Act 3) has been folded into Act 2 as delegate 2
-(`mfg-knights/CriteriaBasedSearchAndFilter`) — the CBSF delegate owns its own pref enablement +
-PSL assignment as its first steps, so no separate Act is needed.
-
-When a new Act's sub-skill is added under `source/skills/<act-slug>/`, add a row to the ordered
-table above and insert it at the correct position in the **Run Order** below. The dependency rule is
-absolute: **an Act runs only after every Act it depends on has completed successfully.**
-
----
-
-## Run Order (how the orchestrator executes)
-
-Execute the Acts strictly in the order listed in "What this orchestrator runs today", top to
-bottom. For each Act:
-
-1. **Preflight the shared runtime once, up front.** Resolve and confirm the target org and the
-   target user via `dispatch` (see the sub-skill's step 1 — `GET /services/data/` for API version,
-   `GET /services/data/v<apiVersion>/chatter/users/me` for user identity). Capture `username`,
-   `apiVersion`, `userId`. If the identity call fails with an auth error, STOP and ask the user to
-   re-authenticate the MCP server connection before running any Act. Pass this resolved context
-   to every Act so no Act re-resolves it.
-2. **Delegate to the Act's sub-skill.** Hand off to the sub-skill for that Act and let it own its
-   own workflow, verification, and reporting. Do not re-implement an Act's mechanics in this
-   orchestrator — the sub-skill is the single source of truth for how that Act is performed.
-   Follow the sub-skill's instructions exactly as written, including any interactive pauses
-   (e.g. the GenAI UI handoff in Act 0 step 6 — this is a deliberate HARD STOP that waits for
-   user confirmation before resuming; it is NOT a failure or block).
-3. **Gate on the Act's outcome before continuing.**
-   - **Success** → proceed to the next Act.
-   - **Partial success with reported gaps** (e.g. a license-gated preference reads back `false`
-     after write — such as Einstein Copilot or Industries Field Service) → the sub-skill will report
-     the gap and continue. This is NOT a block — proceed to the next Act.
-   - **Blocked** (a prerequisite the Act cannot satisfy headlessly AND that downstream Acts depend
-     on — e.g. the Einstein GPT Platform PSL missing, which blocks Agent Platform) → **STOP the
-     ordered run.** Do not start a dependent Act on top of an incomplete foundation. Report exactly
-     what blocked it, with the failing API response, and what a later Act now cannot rely on.
-   - **Hard failure** → STOP, report, do not continue.
-4. **Accumulate results** for the consolidated report (see Output Expectations).
-
-Because every Act is idempotent, re-running the orchestrator after fixing a blocker is always safe —
-already-enabled preferences and already-assigned permission sets are no-ops.
-
----
-
-## Rules / Constraints
-
-| Constraint | Rationale |
-|---|---|
-| Foundation + agent creation + DPE grounding only | This orchestrator runs enablement Acts (preferences + permission sets), agent lifecycle Acts (create + activate from NGA template), and DPE-grounding Acts (clone + activate the DPE definitions the agent needs, via delegation). It never creates business data, **runs** Data Pipelines jobs, registers `SearchableObjDataSyncInfo`, or configures messaging channels — those are separate downstream skills. |
-| Fixed, dependency-respecting order | Later Acts assume Act 0's foundation is in place. Running out of order (or continuing past a blocked Act) produces a half-configured org that fails silently later. |
-| Delegate, don't duplicate | Each Act's mechanics live in its sub-skill. The orchestrator sequences and reports; it does not re-encode deploy payloads or REST calls. |
-| Resolve shared runtime once, pass it down | Resolve org / user / API version at the orchestrator level and hand it to every Act. No Act should re-resolve or hardcode it. |
-| Headless only (one documented exception) | No browser, no manual Setup clicks — same rule every Act follows, with exactly one exception: the Automotive Generative AI legal-consent preference (Act 0, step 6) requires a UI handoff where the skill STOPS and WAITS for the user to confirm they've toggled it on in Setup. If nothing else headless works, STOP and explain. |
-| Stop on block, never cascade a failure | A dependent Act must not start on an incomplete foundation. Halt the run and report. |
-
----
-
-## Adding a new Act (contributor note)
-
-To extend this orchestrator with another setup Act:
-
-1. Author the new Act as a sub-skill at
-   `contributed-skills/auto-avengers/automotive-sales-concierge-foundation-orchestrator/source/skills/<act-slug>/SKILL.md`
-   (same shape as `enable-automotive-sales-concierge-foundation/`).
-2. Add it to the ordered tables above, and insert it at the correct dependency position in the Run
-   Order.
-3. Re-run the skill-onboarding harness. The harness treats each sub-skill as its own delegating
-   sibling and refreshes this parent's delegation edges automatically — you do not hand-wire the
-   parent-to-child link.
-
-Keep each Act narrow and setup-scoped. If a capability needs data creation, pipeline execution, or
-agent deployment, it is not an Act of this orchestrator — it belongs to a downstream skill.
-
----
-
-## Output Expectations
-
-Report, in order:
-1. **Org** — alias, username, API version (resolved once, shared across Acts).
-2. **Run plan** — the ordered list of Acts to run this invocation (today: Act 0 → Act 1 → Act 2).
-3. **Per-Act result** — for each Act, its own before/after summary and status
-   (Success / Blocked / Failed).
-4. **Consolidated foundation state** — is the Automotive Sales Concierge foundation fully enabled?
-   If not, which Act stopped the run and why (with the specific blocker named).
-5. **Next step** — if fully enabled, the setup foundation is ready for the downstream (non-setup)
-   Acts; if blocked, the exact entitlement/license the user must obtain before re-running.
-
----
-
-## Delegated Sub-Skills
-
-| Act | Sub-skill directory | Read for details |
-|---|---|---|
-| Act 0 — Enable the Foundation | `source/skills/enable-automotive-sales-concierge-foundation/` | That sub-skill's `SKILL.md` (fully self-contained) |
-| Act 1 — Create & Activate the Agent | `source/skills/create-automotive-sales-concierge-agent/` | That sub-skill's `SKILL.md` (fully self-contained) |
-| Act 2 — Ground the Agent's Search Infrastructure | `source/skills/ground-agent-data-pipelines/` | That sub-skill's `SKILL.md` (thin delegator) + its `constants.json`; delegates sequentially to `mfg-spartans/dpe-clone-activate` (DPE), `mfg-knights/CriteriaBasedSearchAndFilter` (CBSF config), and `mfg-knights/activate-flexipage-via-action-override` (FlexiPage deploy) |
-
-<!-- skill-validate: ignore-start -->
-<!-- codey:operations start -->
-## Operations Reference
-
-Operations this skill uses — routing and dependency reference.
-
-| Operation | Purpose | Status | Call | Depends on |
-|-----------|---------|--------|------|------------|
-| `preflight-api-version` | read | — | `GET /services/data/` | — |
-| `preflight-resolve-user` | read | — | `GET /services/data/v<apiVersion>` | `preflight-api-version` |
-| `delegate-automotive-sales-concierge-foundation-orchestrator-enable-automotive-sales-concierge-foundation` | delegate | — | → `auto-avengers/automotive-sales-concierge-foundation-orchestrator-enable-automotive-sales-concierge-foundation` | `preflight-resolve-user` |
-| `delegate-automotive-sales-concierge-foundation-orchestrator-create-automotive-sales-concierge-agent` | delegate | — | → `auto-avengers/automotive-sales-concierge-foundation-orchestrator-create-automotive-sales-concierge-agent` | `delegate-automotive-sales-concierge-foundation-orchestrator-enable-automotive-sales-concierge-foundation` |
-| `delegate-automotive-sales-concierge-foundation-orchestrator-ground-agent-data-pipelines` | delegate | — | → `auto-avengers/automotive-sales-concierge-foundation-orchestrator-ground-agent-data-pipelines` | `delegate-automotive-sales-concierge-foundation-orchestrator-create-automotive-sales-concierge-agent` |
-<!-- codey:operations end -->
-<!-- skill-validate: ignore-end -->

package/skills/automotive-cloud-sales-concierge-orchestrator-agent-create/SKILL.md

@@ -1,360 +0,0 @@
----
-name: automotive-cloud-sales-concierge-orchestrator-agent-create
-description: "Create and activate the Automotive Sales Concierge agent from the pre-installed NextGen Authoring (NGA) template in a live org — entirely headless and API-first, driven only through the Headless 360 MCP server's dispatch tool. This is \"Act 1\" (Create & Activate the Agent) of the Automotive Sales Concierge steel thread. It resolves the AutomotiveSalesConcierge NGA template script, creates an NGA bundle from it, publishes the bundle (materializing the BotDefinition + BotVersion), activates the bundle version, and verifies the runtime BotVersion reached Active status. TRIGGER when the user wants to \"create the Automotive Sales Concierge agent\", \"deploy the automotive agent\", \"activate the sales concierge agent\", \"create the Industries Sales Assistance agent with the Automotive Sales Concierge subagent\", or otherwise asks to stand up the agent from the NGA template. DO NOT TRIGGER when the user wants to enable org preferences or assign permission sets (that's Act 0), create business data, run Data Pipelines jobs, or configure messaging channels — those are separate Acts served by other skills."
-metadata:
-  version: "1.0"
----
-
-# Create & Activate Automotive Sales Concierge Agent (Act 1)
-
-Create the **Automotive Sales Concierge** agent from the pre-installed **NextGen Authoring (NGA)**
-template, publish and **activate** the agent — directly in a live org, **headlessly and API-first,
-using only the Headless 360 MCP server's `dispatch` tool**.
-
-This is **Act 1 — "Create & Activate the Agent"** of the Automotive Sales Concierge steel thread.
-It runs **after Act 0** (Enable the Foundation) has completed successfully — Act 0 turns on the
-Agentforce platform prerequisites and Automotive Cloud org preferences that this Act depends on.
-
-The NGA template `AutoGenAi__AutomotiveSalesConcierge` (type: `AgentforceEmployeeAgent`) ships with
-the Automotive Cloud package. It contains the full agent definition in afscript format — the router,
-topics, and the **AutomotiveSalesConcierge** subagent with all 13 flow/standard actions
-pre-configured. The template is resolved by name, then the script is used to create an NGA bundle
-which is published and activated through the **native NextGen Authoring Connect API**.
-
----
-
-## The command surface — `dispatch` ONLY (no sf CLI, no curl, no browser)
-
-Every call this skill makes is a single MCP tool call:
-
-```text
-dispatch(method, path, body?, query_params?)
-```
-
-- The MCP server resolves the org, instance URL, and bearer token on its own. **Do not** ask for or
-  handle tokens, instance URLs, org IDs, or API versions as inputs — `dispatch` supplies them.
-- Use the API version resolved by the orchestrator's preflight (passed down as shared runtime context).
-
-> **Absolute rules (inherited from the Headless 360 factory):**
-> - **`sf` CLI must NEVER appear as a step here.** Hosted-MCP customers have no shell.
-> - **No browser / no manual Setup click.** Every step in this Act is fully headless.
-
----
-
-## Prerequisites (verified by the orchestrator before invoking this Act)
-
-This Act assumes the orchestrator has already:
-1. Resolved the shared runtime (org, userId, apiVersion) in its preflight.
-2. Completed Act 0 successfully — specifically:
-   - `EinsteinGPTPlatformEnabled` = true (pref #1)
-   - `AgentPlatformEnabled` = true (pref #2)
-   - `EinsteinGPTCopilotEnabled` = true (pref #3, or reported as license-level gap)
-   - `IndustriesAutomotiveEnabled` = true (pref #4)
-   - `AutomotiveAgentsEnabled` = true (pref #6)
-   - `IndustriesAutomotiveGenerativeAIEnabled` = true (the UI-only pref — gated on legal consent)
-   - Foundation permission sets assigned (including Agentforce Default Agent access)
-3. NextGen Authoring is provisioned in the org (NGA license active).
-
-If any Agentforce platform prerequisite is missing, this Act will fail. The orchestrator gates on
-Act 0 success before invoking Act 1.
-
----
-
-## Scope — AGENT CREATION & ACTIVATION ONLY (NextGen Authoring path)
-
-- **In scope**
-  - Probing NGA enablement (the `/nextgen-authoring/bundles` endpoint must return 200).
-  - Resolving the `AutoGenAi__AutomotiveSalesConcierge` template's `agentScript` via the
-    agent-authoring catalog (`get-agent-templates`).
-  - Checking name idempotency (is a bundle with this name already created?).
-  - Creating an NGA bundle + draft version from the resolved script.
-  - Publishing the bundle version (materializes BotDefinition + BotVersion).
-  - Activating the bundle version (promotes BotVersion to Active).
-  - Verifying the runtime `BotVersion.Status == "Active"`.
-
-- **Out of scope** (delegate — do NOT attempt here)
-  - Enabling org preferences or assigning permission sets (Act 0).
-  - Creating business data (accounts, vehicles, opportunities, etc.).
-  - Configuring messaging channels, Experience Cloud sites, or deployment surfaces.
-  - Running Data Pipelines jobs.
-  - Modifying the agent's topics, actions, or instructions post-creation (the template is complete).
-  - Using the legacy Copilot Builder `createCopilotFromTemplate` path (that is NOT NextGen Authoring).
-
----
-
-## Required Inputs
-
-All provided by the orchestrator's shared runtime context:
-
-- **`<apiVersion>`** — the org's max API version (from orchestrator preflight step 1). NGA requires v64+.
-- **`<userId>`** — the authenticated user's Id (from orchestrator preflight).
-
-Template name is fixed: **`AutoGenAi__AutomotiveSalesConcierge`** (the fully-qualified NGA template name).
-
----
-
-## The 13 expected actions in the template
-
-The `AutomotiveSalesConcierge` agent template (defined in the `.agent` afscript file) includes
-these actions. After activation, the verify step confirms the agent reached Active status:
-
-| # | Action Name | Target | Purpose |
-|---|---|---|---|
-| 1 | SearchVehicleDefinitions | `flow://AutoGenAi__SearchVehicleDefinitions` | Search vehicle definitions by NLP query |
-| 2 | GetDealersForSrchCriteria | `flow://AutoGenAi__GetDealersForSrchCriteria` | Find dealers by location/criteria |
-| 3 | CreaOppAndPrefSell | `flow://IndSalesConcierge__CreaOppAndPrefSell` | Create opportunity + preferred seller |
-| 4 | CreateQuoteFromOppt | `flow://IndSalesConcierge__CreateQuoteFromOppt` | Create quote from opportunity |
-| 5 | GetAvailableAppointmentSlots | `standardInvocableAction://getAvailableAppointmentSlots` | Get test drive time slots |
-| 6 | ScheduleAutomotiveAppointment | `standardInvocableAction://scheduleSvcApptActnblEvntOrch` | Book test drive appointment |
-| 7 | GetVehiclesForCustomer | `flow://AutoGenAi__GetVehiclesForACustomer` | Get customer's existing vehicles |
-| 8 | CreateVehicleAppraisal | `flow://IndAppraisals__CreateVehicleAppraisal` | Initiate trade-in appraisal |
-| 9 | SummarizeCustomerAccount | `generatePromptResponse://AutoGenAi__SummarizeCustomerAccount` | Summarize customer account |
-| 10 | QueryRecords | `standardInvocableAction://queryRecords` | SOQL via natural language |
-| 11 | SummarizeRecord | `standardInvocableAction://getRecordSummarizationPrompt` | Summarize any record |
-| 12 | DraftOrReviseEmail | `standardInvocableAction://getDraftOrReviseEmailPrompt` | Draft/revise emails |
-| 13 | SrchPrdctAndParts | `flow://IndSalesConcierge__SrchPrdctAndParts` | Search accessories/related parts |
-
----
-
-## Workflow — NextGen Authoring: resolve → create → publish → activate → verify
-
-Follow in order. Report progress per step.
-
-### Step 1. Resolve API version
-
-```text
-dispatch("GET", "/services/data/")
-```
-
-Take `max(versions[].version)` → `{v}`. NGA requires v64 floor; use the org's newest version.
-
-### Step 2. Probe NGA enablement
-
-```text
-dispatch("GET", "/services/data/v<apiVersion>/nextgen-authoring/bundles")
-```
-
-- **200** = NGA is live and the user has read access. Proceed.
-- **403** = NGA is NOT provisioned or the user lacks `userCanAccessAuthoringBundle`. **STOP.**
-  Report that NextGen Authoring is not enabled in this org; this is a blocker.
-
-This also lists any already-created bundles — check if `AutomotiveSalesConcierge` already exists.
-
-### Step 3. Resolve template script (the one off-native step)
-
-The native NextGen Authoring API has no template catalog. Template name → agentScript resolution
-goes through the headless agent-authoring dispatch route:
-
-```text
-dispatch("GET", "/headless/invoke/platform/agent-authoring/get-agent-templates")
-```
-
-The dispatcher auto-resolves the route and returns the unwrapped response directly.
-
-**Response:** array of template entries. Each has `{devName, namespace, name (FQN = namespace__devName),
-agentType, agentScript, templateLabel, templateDescription, iconUrl, ...}`.
-
-**Match logic:** find the entry where `name` is EXACTLY equal to `AutoGenAi__AutomotiveSalesConcierge`
-(strict equality — NOT `startsWith`, NOT `includes`, NOT substring match). There are multiple
-templates with similar prefixes (e.g. `AutoGenAi__AutomotiveSalesConciergeCrossChannel`); you MUST
-match on exact string equality only. If exact FQN match yields no result, fall back to exact
-`devName` == `AutomotiveSalesConcierge` (again, strict equality — not partial match).
-
-**Capture:** the matching entry's `agentScript` (the full afscript — typically ~60KB+).
-
-**If no template found:** STOP. Report that the `AutoGenAi__AutomotiveSalesConcierge` NGA template
-is not installed in this org. The Automotive Cloud package may not be deployed or the
-`IndustriesAutomotiveGenerativeAIEnabled` preference is off.
-
-### Step 4. Check name idempotency
-
-Before creating, check if a bundle with this name already exists:
-
-```text
-dispatch("GET", "/services/data/v<apiVersion>/nextgen-authoring/bundles/names")
-```
-
-Check if `AutomotiveSalesConcierge` (or whatever `apiName` you intend to use) is already taken.
-
-**If already exists:** The agent was already created in a prior run. Query for its bundle version
-and check its status — if already Active, report success and skip to verify. If in DRAFT state,
-proceed to publish + activate.
-
-### Step 5. Create NGA bundle from template script
-
-**CRITICAL — rewrite `config.developer_name` and `config.agent_label` in the agentScript** to a
-run-scoped unique value before creating. The collision key for `publish` is the script's internal
-`developer_name`, NOT the bundle `apiName`. If a BotDefinition with the same `developer_name`
-already exists, publish will fail with `400 DUPLICATE_VALUE`.
-
-For the Automotive Sales Concierge, use:
-- `config.developer_name`: `AutomotiveSalesConcierge` (or a run-scoped variant if re-running)
-- `config.agent_label`: `Automotive Sales Concierge`
-
-**DO NOT change `config.agent_template`** — that is the template-lineage pointer and must stay as-is.
-
-```text
-dispatch("POST", "/services/data/v<apiVersion>/nextgen-authoring/bundles",
-  body={
-    "apiName": "AutomotiveSalesConcierge",
-    "label": "Automotive Sales Concierge",
-    "assets": [
-      {
-        "resourceName": "definition.agent",
-        "resourceType": "agentDefinition",
-        "resourceContent": "<the resolved agentScript from Step 3>",
-        "resourceAuthoringFormat": "afscript",
-        "authoringFormatVersion": "v1.0"
-      }
-    ]
-  })
-```
-
-**Response:** `{bundleId, id (= bundleVersionId), apiName, label, versionStatus: "DRAFT", ...}`.
-
-Capture **`bundleId`** and **`bundleVersionId`** (the `id` field). Thread these through remaining steps.
-
-**Error handling:**
-- Auth/permission errors → user lacks `userCanEditAuthoringBundle`. Report the exact error.
-- 500 on large script → fallback: create-empty bundle then PATCH `updateBundleVersion` to attach
-  the asset separately.
-
-### Step 6. Publish the bundle version
-
-```text
-dispatch("POST", "/services/data/v<apiVersion>/nextgen-authoring/bundle-versions/<bundleVersionId>/publish")
-```
-
-**Response** carries:
-- `publishedBotId` — the materialized BotDefinition Id
-- `publishedBotVersionId` — the materialized BotVersion Id
-- `versionStatus: "Published"`
-
-Capture **`publishedBotId`** and **`publishedBotVersionId`** for verification.
-
-**Error handling:**
-- `400 DUPLICATE_VALUE: BotDefinition duplicates value on record with id: <developer_name>` →
-  the script's `config.developer_name` collides with an existing BotDefinition. Either the agent
-  already exists (check and proceed to activate the existing one) or use a unique developer_name.
-- Permission errors → user lacks `userCanEditAuthoringBundle`.
-
-### Step 7. Activate the bundle version
-
-```text
-dispatch("POST", "/services/data/v<apiVersion>/nextgen-authoring/bundle-versions/<bundleVersionId>/activate")
-```
-
-**Returns HTTP 204** (empty body). Activation is **async** — do NOT expect the status to flip
-immediately.
-
-**CRITICAL — activation is NOT blind-retry-safe.** A second activate POST while the first is still
-in flight returns 500 UNKNOWN_EXCEPTION. On timeout, **read status back before retrying** — never
-blind-retry.
-
-### Step 8. Verify activation
-
-Two-level verification:
-
-**8a. Accepted-checkpoint (async activation queued):**
-
-```text
-dispatch("GET", "/services/data/v<apiVersion>/nextgen-authoring/bundle-versions/<bundleVersionId>")
-```
-
-Assert `targetStatus == "ACTIVE"`. This confirms the activation was accepted.
-Note: `versionStatus` may still show `"PUBLISHED"` — that is normal (it does NOT update to ACTIVE
-synchronously). Check `targetStatus`, not `versionStatus`.
-
-**8b. Authoritative Active proof (runtime record):**
-
-```text
-dispatch("GET", "/services/data/v<apiVersion>/query",
-  query_params={"q": "SELECT Id, Status FROM BotVersion WHERE Id = '<publishedBotVersionId>'"})
-```
-
-Assert `Status == "Active"`.
-
-If not Active yet, wait 5-10 seconds and re-query (activation is async). Up to 3 retries with
-increasing delays (5s, 10s, 15s).
-
----
-
-## Findings — encode as invariants
-
-1. **Enablement gate (step 2).** 403 on the read probe = NGA not provisioned → flag, don't retry.
-2. **Name→script is the only off-native step and keys on the FQN.** Native NGA Connect has no
-   template catalog. Match `templateName` against `namespace__devName`.
-3. **Publish is a distinct step before activate.** create → publish → activate. Activating a
-   DRAFT/unpublished version fails.
-4. **Activation is async AND not blind-retry-safe.** Returns 204; a 2nd activate while the first
-   is in flight returns 500. Read status back before retrying.
-5. **Verify via the RUNTIME record, not authoring versionStatus.** Use `targetStatus == "ACTIVE"`
-   for the accepted-checkpoint and runtime `BotVersion.Status == "Active"` for the hard pass.
-6. **`validateAgentJson` is a known false-negative — do NOT gate publish on it.**
-7. **Idempotency's collision key is the script's `config.developer_name`, NOT the bundle apiName.**
-   `createBundleWithVersion` auto-suffixes a colliding apiName, but publish materializes a
-   BotDefinition keyed on developer_name. On collision, publish fails with DUPLICATE_VALUE.
-8. **The ~60KB+ script must round-trip inline.** There is no name→instantiate server API; create
-   takes the full agentScript inline in the request body.
-
----
-
-## Rules / Constraints
-
-| Constraint | Rationale |
-|---|---|
-| `dispatch` only — no `sf` CLI, no `curl`, no browser | Hosted-MCP customers have no shell. |
-| Use the **NextGen Authoring Connect API** (`/nextgen-authoring/...`) | This is the canonical NGA path; do NOT use the legacy Copilot Builder `createCopilotFromTemplate`. |
-| Template is `AutoGenAi__AutomotiveSalesConcierge` (FQN) | This is the NextGen template shipped in Core; NOT the legacy `IndSalesConcierge__IndustriesSalesAssistance` copilot template. |
-| Publish BEFORE activate | Create → Publish → Activate is the mandatory sequence. |
-| Never blind-retry activate | Activation is async; read status back before retrying. |
-| No hardcoded IDs | Resolve bundleId, bundleVersionId, publishedBotId, publishedBotVersionId at runtime. |
-| Do not modify the agent post-creation | The template afscript is the source of truth. |
-| Do not delete the agent | Leave it in place for validation by the admin. |
-
----
-
-## Output Expectations
-
-Report, in order:
-1. **NGA enablement** — confirmed NextGen Authoring is provisioned (step 2 returned 200).
-2. **Template resolution** — the `AutoGenAi__AutomotiveSalesConcierge` template's agentScript
-   resolved successfully (size in chars).
-3. **Bundle creation** — `bundleId`, `bundleVersionId`, `apiName`, `versionStatus: DRAFT`.
-4. **Publish** — `publishedBotId`, `publishedBotVersionId`, `versionStatus: Published`.
-5. **Activation** — activation accepted (204), `targetStatus: ACTIVE`.
-6. **Verification** — runtime `BotVersion.Status == "Active"` confirmed.
-7. **Summary** — the Automotive Sales Concierge agent is live. Provide:
-   - `bundleId`, `bundleVersionId` (NGA identifiers)
-   - `publishedBotId` (BotDefinition Id / copilotId)
-   - `publishedBotVersionId` (BotVersion Id)
-   - `apiName` / `DeveloperName`
-8. **Next step** — hand back to the orchestrator for the next Act.
-
----
-
-## Troubleshooting
-
-| Symptom | Cause | Resolution |
-|---|---|---|
-| Step 2 returns 403 | NGA not provisioned or user lacks NGA perms | Verify NGA license is provisioned and user has Agent Authoring permissions. |
-| Template not found in `get-agent-templates` | Automotive Cloud package not deployed, or `IndustriesAutomotiveGenerativeAIEnabled` pref is off | Verify Act 0 completed; check the UI-only GenAI preference. |
-| `createBundleWithVersion` 500 on large script | Intermittent server issue with large payloads | Fallback: create-empty then PATCH `updateBundleVersion` to attach asset. |
-| Publish fails with `DUPLICATE_VALUE: BotDefinition duplicates value` | A BotDefinition with same `developer_name` already exists | Check existing agent; if Active, report success. If stale, use a unique developer_name. |
-| Activate returns 500 UNKNOWN_EXCEPTION | Double-activate race condition | Read `getBundleVersion` status; if `targetStatus == "ACTIVE"`, the first activate succeeded. |
-| `BotVersion.Status` not Active after activate 204 | Activation is async | Wait 5-15s and re-query. Up to 3 retries. |
-| `validateAgentJson` reports errors on valid script | Known false-negative (YAML parser vs afscript) | Ignore validate failures; proceed to publish (the real compiler). |
-
----
-
-## Cross-Skill Integration
-
-This skill is **Act 1** and is designed to be invoked by the **Automotive Sales Concierge Foundation
-Orchestrator** (its parent) after Act 0 succeeds.
-
-| Need | Belongs to |
-|---|---|
-| Enable org preferences + assign permission sets | Act 0 — `enable-automotive-sales-concierge-foundation` |
-| Configure messaging channels / deployment surfaces | later Act — channel deployment |
-| Create business data (accounts, vehicles, opportunities) | later Act — data setup |
-| Run Data Pipelines jobs | later Act — DPE enablement |