Add Slack connector input targets and element template guidance

.claude/skills/new-agent/SKILL.md

@@ -24,6 +24,6 @@ Tools can also be **sub-processes** for multi-step async patterns:
 - Use this for send+wait, request+callback, or any tool needing async interaction
 - See the "Sub-flow Tool Pattern" section in camunda-dev-guide.md
 
-For non-REST connectors as tools, look up their task type and input targets via the Camunda Docs MCP tool.
+For non-REST connectors as tools, look up their **element template JSON** from `https://github.com/camunda/connectors/tree/main/connectors/` for exact input target names. Do NOT add `zeebe:modelerTemplate` attributes or `elementTemplateId` headers — these cause errors in Web Modeler. The user applies templates from the Modeler UI after upload.
 
 Save to resources/. Deploy with c8ctl.

.claude/skills/new-process/SKILL.md

@@ -6,7 +6,16 @@ argument-hint: "[process description]"
 
 Generate a deployable Camunda 8 BPMN file. Find and read `camunda-dev-guide.md` for Zeebe-specific patterns. For agentic processes with AI agents and ad-hoc sub-processes, use `/new-agent` instead.
 
-If the process uses non-REST connectors (Slack, Kafka, SendGrid, etc.), look up their task type and input targets via the Camunda Docs MCP tool or at https://docs.camunda.io/docs/components/connectors/out-of-the-box-connectors/
+If the process uses non-REST connectors (Slack, Kafka, SendGrid, etc.), look up their **element template JSON** from `https://github.com/camunda/connectors/tree/main/connectors/` to get the exact input target names, required fields, and conditional properties. Fall back to Camunda Docs MCP or https://docs.camunda.io/docs/components/connectors/out-of-the-box-connectors/
+
+## Element Templates — IMPORTANT
+
+Do NOT manually add `zeebe:modelerTemplate` attributes or `elementTemplateId`/`elementTemplateVersion` task headers in generated BPMN XML. These cause "Template Not found" errors in Web Modeler.
+
+Instead:
+- Generate BPMN with correct `zeebe:taskDefinition`, `zeebe:ioMapping`, and `zeebe:taskHeaders` matching the template's property bindings
+- Use the official element template JSON to determine correct input target names (they differ per connector — some use flat names, others use `data.*` or `authentication.*` prefixes)
+- After uploading to Web Modeler, the user applies the connector template from the Modeler UI
 
 $ARGUMENTS
 

docs/camunda-dev-guide.md

@@ -128,10 +128,13 @@ All outbound connectors (REST, Slack, Kafka, SendGrid, OpenAI, AWS, etc.) use th
 ```
 
 ### To configure any specific connector:
-1. Look up the connector in Camunda docs (use the Docs MCP tool or https://docs.camunda.io/docs/components/connectors/out-of-the-box-connectors/)
-2. Find the **task type** (e.g., `io.camunda:http-json:1`, `io.camunda:slack:1`, `io.camunda:kafka:1`)
-3. Find the **required input target names** and expected values
-4. Apply them to the pattern above
+1. **Fetch the element template JSON** from `https://github.com/camunda/connectors/tree/main/connectors/<name>/element-templates/` — this is the authoritative source for input target names
+2. Find the **task type** in the template (property with `"binding": {"property": "type", "type": "zeebe:taskDefinition"}`)
+3. Find all **input bindings** (properties with `"binding": {"name": "...", "type": "zeebe:input"}`) — the `name` field is the exact `target` value for `<zeebe:input>`
+4. Note **conditional properties** — some inputs are only required when another property has a specific value (e.g., `data.text` only when `method` = `chat.postMessage`)
+5. Apply them to the universal pattern above
+
+Fall back to Camunda Docs MCP or https://docs.camunda.io/docs/components/connectors/out-of-the-box-connectors/ if the template JSON is unavailable.
 
 Secrets (API keys, tokens) are referenced as `{{secrets.MY_SECRET}}` in input values.
 
@@ -140,6 +143,7 @@ Secrets (API keys, tokens) are referenced as `{{secrets.MY_SECRET}}` in input va
 - Outputs (result extraction) go in `<zeebe:taskHeaders>` as `<zeebe:header>` — NOT in ioMapping
 - Response shape for REST: `response.status`, `response.headers`, `response.body`
 - `retryBackoff` is a reserved header key handled by the connector runtime
+- **Input target naming is connector-specific**: REST uses flat names (`method`, `url`, `body`) + `authentication.*`; Slack uses `data.*` prefix for method-specific fields. Always check the element template — do NOT guess target names.
 
 ### REST connector example (most common, `io.camunda:http-json:1`):
 ```xml
@@ -161,6 +165,61 @@ Secrets (API keys, tokens) are referenced as `{{secrets.MY_SECRET}}` in input va
 ### REST connector input targets (must be exact):
 `method`, `url`, `body`, `headers`, `queryParameters`, `authentication.type`, `authentication.token`, `authentication.username`, `authentication.password`, `connectionTimeoutInSeconds`, `readTimeoutInSeconds`
 
+### Slack connector (`io.camunda:slack:1`):
+
+Input targets vary by method. Root-level: `token`, `method`. Method-specific fields use `data.*` prefix.
+
+**chat.postMessage:**
+```xml
+<zeebe:input source="{{secrets.SLACK_BOT_TOKEN}}" target="token"/>
+<zeebe:input source="chat.postMessage" target="method"/>
+<zeebe:input source="plainText" target="data.messageType"/>
+<zeebe:input source="=slackChannel" target="data.channel"/>
+<zeebe:input source="=slackMessage" target="data.text"/>
+```
+`data.messageType` is required (`plainText` or `messageBlock`). Channel accepts channel name, user ID (for DMs), or email.
+
+**conversations.create:**
+```xml
+<zeebe:input source="conversations.create" target="method"/>
+<zeebe:input source="=channelName" target="data.newChannelName"/>
+<zeebe:input source="PUBLIC" target="data.visibility"/>
+```
+Note: field is `data.newChannelName` (not `data.name`), `data.visibility` is `PUBLIC`/`PRIVATE` (not boolean).
+
+**conversations.invite:**
+```xml
+<zeebe:input source="conversations.invite" target="method"/>
+<zeebe:input source="=channelId" target="data.channelId"/>
+<zeebe:input source="=userIds" target="data.users"/>
+```
+
+Other methods: `reactions.add`, `pins.add`, `pins.remove`. Look up the element template JSON for full bindings.
+
+### Element Templates — do NOT set manually in BPMN XML
+
+Element templates are JSON files that define the connector's properties, UI groups, and validation rules for Web Modeler. Each connector has one in the `camunda/connectors` repo.
+
+**When generating BPMN programmatically or with AI:**
+
+Do NOT add `zeebe:modelerTemplate` / `zeebe:modelerTemplateVersion` attributes on service tasks, and do NOT add `elementTemplateId` / `elementTemplateVersion` task headers. These cause "Template Not found" errors in Web Modeler because the template must be resolved from the Modeler's internal catalog — raw XML references don't link to it.
+
+Instead:
+1. Generate BPMN with correct `zeebe:taskDefinition`, `zeebe:ioMapping`, and `zeebe:taskHeaders` that match the template's property bindings
+2. The user applies the connector template from the Web Modeler UI after upload (right-click task → "Apply template" or use the properties panel)
+3. Web Modeler will recognize the existing inputs and map them to the template fields
+
+**Reading element template JSON — how to extract input targets:**
+
+Each template property has a `binding` object:
+- `"type": "zeebe:input"` with `"name": "..."` → the `name` is the exact `target` for `<zeebe:input>`
+- `"type": "zeebe:taskDefinition"` with `"property": "type"` → the task type
+- `"type": "zeebe:taskHeader"` with `"key": "..."` → the `key` for `<zeebe:header>`
+- Properties with `"condition"` are only required when the condition matches (e.g., `body` only for POST/PUT/PATCH)
+- Properties with `"value"` have defaults — include them if they're required (`constraints.notEmpty`)
+
+Template JSON location: `https://github.com/camunda/connectors/tree/main/connectors/<name>/element-templates/`
+
 ---
 
 ## Agentic BPMN (AI Agents)