Salesforce Data Cloud Connect phase. TRIGGER when: user manages Data Cloud connections, connectors, connector metadata, tests a connection, browses source objects or databases, or sets up a new source system. DO NOT TRIGGER when: the task is about data streams or DLOs (use sf-datacloud-prepare), DMO

sf datacloud connect: Data Cloud Connect Phase Use this skill when the user needs source connection work : connector discovery, connection metadata, connection testing, source object browsing, connector schema inspection, or connector specific setup payloads for external sources. When This Skill Owns the Task Use sf datacloud connect when the work involves: sf data360 connection connector catalog inspection connection creation, update, test, or delete browsing source objects, fields, databases, or schemas identifying connector types already in use preparing connector definitions for Snowflake, SharePoint Unstructured, or Ingestion API sources Delegate elsewhere when the user is: creating data streams or DLOs → [sf datacloud prepare](../sf datacloud prepare/SKILL.md) creating DMOs, mappings, IR rulesets, or data graphs → [sf datacloud harmonize](../sf datacloud harmonize/SKILL.md) writing Data Cloud SQL or search index workflows → [sf datacloud retrieve](../sf datacloud retrieve/SKILL.md) Required Context to Gather First Ask for or infer: target org alias connector type or source system whether the user wants inspection only or live mutation connection name or ID if one already exists whether credentials are already configured outside the CLI whether the user also expects stream creation right after connection setup whether the source is a database, an unstructured document source, or an Ingestion API feed Core Operating Rules Verify the plugin runtime first; see [../sf datacloud/references/plugin setup.md](../sf datacloud/references/plugin setup.md). Run the shared readiness classifier before mutating connections: node ~/.claude/skills/sf datacloud/scripts/diagnose org.mjs o <org phase connect json . Prefer read only discovery before connection creation. Suppress linked plugin warning noise with 2 /dev/null for standard usage. Remember that connection list requires connector type . For connection test , pass connector type when resolving a non Salesforce connection by name. Discover existing connector types from streams first when the org is unfamiliar. Use curated example payloads before inventing connector specific credentials or parameters. For connector types outside the curated examples, inspect a known good UI created connection via REST before building JSON. Do not promise API based stream creation for every connector type just because connection creation succeeds. Recommended Workflow 1. Classify readiness for connect work 2. Discover connector types 3. Inspect connections by type 4. Inspect a specific connection or uploaded schema 5. Test or create only after discovery 6. Start from curated example payloads for external connectors Use the phase owned examples before inventing a payload from scratch: examples/connections/heroku postgres.json examples/connections/redshift.json examples/connections/sharepoint unstructured.json examples/connections/snowflake connection.json examples/connections/ingest api connection.json examples/connections/ingest api schema.json Typical Ingestion API setup flow: 7. Discover payload fields for unknown connector types Create one in the UI, then inspect it directly: High Signal Gotchas connection list has no true global "list all" mode; query by connector type. The connector catalog name and connection connector type are not always the same label. connection test may need connector type for name resolution when the source is not a default Salesforce connector. An empty connection list usually means "enabled but not configured yet", not "feature disabled". Heroku Postgres, Redshift, Snowflake, SharePoint Unstructured, and Ingestion API all use different credential and parameter shapes; reuse the curated examples instead of guessing. SharePoint Unstructured uses clientId , clientSecret , and tokenEndpoint in the credentials array and does not require a parameters array. Snowflake uses key pair auth and can often be created through the API, but downstream stream creation can still remain UI only. Ingestion API connector setup is incomplete until connection schema upsert has uploaded the object schema. Some external connector credential setup still depends on UI side configuration or external system permissions. Output Format References [README.md](README.md) [examples/connections/heroku postgres.json](examples/connections/heroku postgres.json) [examples/connections/redshift.json](examples/connections/redshift.json) [examples/connections/sharepoint unstructured.json](examples/connections/sharepoint unstructured.json) [examples/connections/snowflake connection.json](examples/connections/snowflake connection.json) [examples/connections/ingest api connection.json](examples/connections/ingest api connection.json) [examples/connections/ingest api schema.json](examples/connections/ingest api schema.json) [../sf datacloud/references/plugin setup.md](../sf datacloud/references/plugin setup.md) [../sf datacloud/references/feature readiness.md](../sf datacloud/references/feature readiness.md) [../sf datacloud/UPSTREAM.md](../sf datacloud/UPSTREAM.md)

sf-datacloud-connect