service-omni-agent-work-sharing-configure

Detect and optionally configure the standard AgentWork object's internal organization-wide default. The supported headless contract is CustomObject:AgentWork metadata: retrieve the current metadata, preserve externalSharingModel, deploy internal sharingModel=Read, and retrieve it again for proof.

This setting grants read visibility broadly to internal users who already have object access. It is not a substitute for an OmniSupervisorConfig: that configuration filters which reps, queues, skills, actions, and tabs a supervisor sees, but does not grant record access.

Inputs

bash scripts/configure-and-report.sh <org-alias> --plan
bash scripts/configure-and-report.sh <org-alias> --confirm-org-wide-visibility

• org-alias is required and must already be authenticated with sf.
• --plan retrieves and reports the current and desired models without deploying.
• --confirm-org-wide-visibility is required only when a run must change Private to Read. Omitting it blocks before deployment.

Safety contract

• A run refuses production customer orgs. Plan mode remains read-only.
• externalSharingModel is preserved exactly.
• Read and ReadWrite already provide internal read visibility and are reused. The skill never narrows ReadWrite to Read.
• The only supported mutation is Private to Read; the skill never grants internal edit access.
• Failed classification, retrieve, deploy, or read-back blocks. It never reports success from an inconclusive response.
• Use a dedicated sandbox or test org because changing an OWD can initiate sharing recalculation.

Workflow

1. Classify the target org and enforce the production-write guard.
2. Retrieve CustomObject:AgentWork through Metadata API.
3. Read internal and external sharing models.
4. Reuse Read or ReadWrite with no deploy.
5. In plan mode, report the proposed Private to Read change and exit yellow.
6. In run mode, require explicit blast-radius confirmation.
7. Deploy only the standard-object metadata containing preserved external sharing and internal Read.
8. Retrieve again and require internal Read plus the unchanged external model.

Output

A single JSON object containing status, previous_sharing_model, sharing_model, external_sharing_model, changed, dry_run, deploy_id, and blocking_issue.

• configured: a confirmed Private to Read deployment succeeded and read-back matched.
• reused: current Read or ReadWrite already provides visibility.
• action_needed: plan mode found Private; no write occurred.
• blocked: a safety guard or supported-API verification failed.

Limitations

• This skill changes visibility, not supervisor scope. Use the supervisor configuration skill for filtering.
• It does not create sharing rules or per-record AgentWorkShare rows.
• It cannot prove a particular supervisor's complete UI access; validate the end-to-end Command Center journey with the intended supervisor user after configuration.

References