service-omni-supervisor-config-deploy

SkillWeb & browsing

Deploys an omni claude skill-style Salesforce Omni-Channel supervisor configuration so agents can set up supervisor monitoring in one step.

Available today. Use it from your connected AI after setup.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the service-omni-supervisor-config-deploy skill

About this skill

Use to stand up a headless Omni-Channel Supervisor Configuration by deploying the classic OmniSupervisorConfig (parent plus supervisor-user and queue companions) in one atomic Metadata API deploy, idempotent via the deploy state. Requires active supervisor users, Omni base settings, and the monitore

What this skill tells your AI

The instructions your AI receives, as published by forcedotcom/sf-skills in skills/service-omni-supervisor-config-deploy/SKILL.md and read by ahel’s review.

Deploy the classic Omni-Channel Supervisor Configuration (OmniSupervisorConfig) to a Salesforce org in a single Metadata API deploy. OmniSupervisorConfig is a full Metadata API type whose XSD embeds every companion as a repeatable child element, so the whole configuration — parent plus companions — lands atomically from one .omniSupervisorConfig-meta.xml file, and Salesforce rolls the companions back together on a partial failure. By default the deploy binds supervisors explicitly by user and lists the queues they monitor; profile scope is optional (see Behavior).

Inputs

bash scripts/deploy-and-report.sh <org-alias> [supervisor_count] [additional_queues_csv] [profiles_csv] [skill_visibility] [config_developer_name]

  • org-alias (required) — must resolve via sf org display --target-org <alias>.
  • supervisor_count (optional, default 1, range 1..5) — how many active supervisor{N}.<suffix>@example.com users to detect and bind.
  • additional_queues_csv (optional) — queue DeveloperNames to bind. When the coordinator supplies this list it is authoritative; when omitted, the canonical CaseQueue, messagingqueue defaults are used. Every requested queue must already exist on the org.
  • profiles_csv (optional) — opt into profile-based fallback scope; supply metadata profile fullNames (e.g. Standard,ServiceCloud). Omit it to bind supervisors by user only.
  • skill_visibility (optional, or OMNI_SUPERVISOR_SKILL_VISIBILITY) — AllSkills or AnySkill. Omitted values preserve an existing config's selection; a new config defaults to AllSkills.
  • config_developer_name (optional, or OMNI_SUPERVISOR_CONFIG_DEVELOPER_NAME; default Omni_Supervisor) — the exact classic OmniSupervisorConfig.DeveloperName to create or reconcile. Supply it when an org has multiple supervisor configurations; the skill never guesses between candidates. OMNI_SUPERVISOR_CONFIG_MASTER_LABEL optionally sets the label for a new config. Existing labels are preserved by default.

Preconditions and safety

  • Target org authenticated via sf CLI, Service Cloud license present, sf CLI ≥ 2.139.6.
  • Active supervisor users exist; their supervisor access is granted separately by service-omni-supervisor-permset-assign.
  • Omni-Channel base settings are enabled (service-omni-base-settings-configure) and the monitored queues exist (service-omni-queue-deploy).
  • Production guardrail: the script computes safe_to_write as IsSandbox OR TrialExpirationDate != null OR OrganizationType in {Developer Edition, Base Edition}, and blocks with no override when it is false. Any unmet precondition returns status: blocked with a blocking_issue and a click-path.

Run

One script does detect, deploy, and verify:

bash scripts/deploy-and-report.sh <org-alias> [supervisor_count] [additional_queues_csv] [profiles_csv] [skill_visibility] [config_developer_name]

It resolves the deploy artifact's tokens from live org state (discovered supervisor users, verified queues, and — only when profiles_csv is given — validated profiles), materializes the selected DeveloperName into the metadata filename and package manifest, then re-queries that same selected config to confirm the parent and its companion rows persisted before emitting JSON to stdout. The coordinator (service-omni-channel-setup-coordinate) invokes it with the authoritative queue list and selected supervisor config; standalone runs use the canonical defaults.

Behavior

Companions. The deployed config always carries omniSupervisorConfigUser (one per supervisor) and omniSupervisorConfigQueue (one per monitored queue). It follows the same single-file Metadata deploy shape as service-omni-service-channel-configure and service-omni-presence-status-deploy.

Profile scope is a fallback, off by default. Core resolves a supervisor's configuration by OmniSupervisorConfigUser first, and only falls back to OmniSupervisorConfigProfile — matched against that supervisor's own profile — when the supervisor has no user row (RealTimeQueueServiceImpl.getSupervisorConfigId). Because this skill binds every supervisor by user, profile companions are unnecessary and are omitted unless you pass profiles_csv. When supplied, values must be metadata fullNames (the <profile> element resolves against the fullName — Standard, ServiceCloud, Admin — not the SOQL User.Profile.Name label); the skill validates each against sf org list metadata -m Profile and blocks on any the org does not expose.

Idempotency and surface preservation. Status follows the Metadata deploy component state: Unchanged → reused, Changed → updated, Created → created. Any other state is a hard failure rather than a silent reused. Before updating an existing parent, the skill snapshots its reference-free OmniSupervisorConfigAction and OmniSupervisorConfigTab rows, validates that each can be recreated safely, then restores and re-queries them after the deploy. A custom/reference-bearing surface blocks before deployment instead of risking silent loss. When reconciling an existing config or verifying preservation, invoke only this skill; do not invoke service-omni-supervisor-surface-deploy, because adding standard rows afterward could mask a preservation failure. The companion surface skill remains responsible only for a separate request to add new standard actions or tabs.

Fail-closed verification. The deploy status must be Succeeded (not SucceededPartial). After the parent config is queryable, the script re-queries the child OmniSupervisorConfigUser and OmniSupervisorConfigQueue rows and requires at least the intended counts — the parent existing does not prove the companions landed. Every read must return a parseable .result.records array; an inconclusive read never counts as zero. Requested queues are validated against ^[A-Za-z0-9_]{1,80}$ before use, and every requested queue must exist or the run blocks (dropping one silently would under-provision the config).

Output contract

A single JSON object to stdout. status is one of created, updated, reused, or blocked. The object also carries config_developer_name, config_id, skill_visibility, surface_preservation snapshot/restoration counts, supervisor_users_bound/supervisor_users_count, queues_bound/queues_count, queue_source (caller_supplied or canonical_default), the re-queried verified_companion_user_count/verified_companion_queue_count, deploy_id, state, manual_actions, and blocking_issue. blocking_issue is null unless status: blocked, in which case it names the missing prerequisite with a click-path in manual_actions. Re-running an unchanged config returns reused and performs no functional change.

Org-side spot-check:

SELECT Id, DeveloperName, MasterLabel FROM OmniSupervisorConfig WHERE DeveloperName='<config_developer_name>'
SELECT COUNT(Id) FROM OmniSupervisorConfigUser  WHERE OmniSupervisorConfigId = :configId
SELECT COUNT(Id) FROM OmniSupervisorConfigQueue WHERE OmniSupervisorConfigId = :configId

Limitations

  • Populates user and queue companions (plus optional profile scope). Existing reference-free Action and Tab rows are preserved across an update; new standard rows are added separately by service-omni-supervisor-surface-deploy. Custom-action, FlexiPage, AWS-dashboard, and AI-agent rows require external references and therefore block an update rather than being reconstructed incompletely.
  • Binds supervisor users discovered by the supervisor{N}.<suffix>@example.com pattern; it does not create them.
  • Does not change AgentWork sharing because that org-wide access control is separate from OmniSupervisorConfig. Use service-omni-agent-work-sharing-configure for supported metadata detection and the explicitly confirmed Private to Read operation.
  • Does not activate the Command Center V2 runtime (a separate settings flip).

References

FileWhen to read
references/api-notes.mdOn a deploy failure — field-by-field OmniSupervisorConfig schema and companion sObject semantics
assets/force-app/main/default/omniSupervisorConfigs/Omni_Supervisor.omniSupervisorConfig-meta.xmlThe token-templated deploy artifact (__MASTER_LABEL__, __SUPERVISOR_USERS_XML__, __PROFILE_SCOPE_XML__, __QUEUE_LIST_XML__, __SKILL_VISIBILITY__); it is materialized under the selected DeveloperName at runtime
assets/package.xml, assets/sfdx-project.jsonManifest and project scaffold materialized into the temp deploy dir

Signals

GitHub stars
1k
Forks
348
Last commit
Sep 2026
Advanced
Item type
skill
Key
service-omni-supervisor-config-deploy
Source
github.com/forcedotcom/sf-skills