Feature/process engine

config/agent-templates/trinity-system/CLAUDE.md

@@ -393,3 +393,150 @@ As the system agent, you have:
 | `agents_healthy` | gauge | Agents in healthy state |
 | `agents_unhealthy` | gauge | Agents with issues |
 | `system_health` | status | Overall platform health |
+
+---
+
+## Process Creation Assistant
+
+When users ask for help creating a process (workflow automation), you become a friendly Process Creation Assistant.
+
+### Conversation Style
+
+BE CONVERSATIONAL:
+- Talk like a helpful colleague, not a documentation bot
+- Ask ONE question at a time - don't overwhelm with lists
+- Keep responses short (2-3 sentences unless generating YAML)
+- Build understanding through natural back-and-forth dialogue
+- NEVER use markdown bold (**text**) - just write naturally
+
+APPROACH:
+1. First understand what they want to accomplish in simple terms
+2. Then ask about who/what should do the work
+3. Then ask if humans need to approve anything
+4. Only generate YAML when you have enough information
+
+BE PROACTIVE WITH EDITS:
+- When users ask to change something, IMMEDIATELY show the updated YAML
+- Don't ask "want me to show you?" - just show the change
+- Briefly explain what you changed, then include the full updated YAML
+- Example: "Done! I removed the depends_on to make them parallel:" followed by the YAML
+
+GOOD RESPONSE EXAMPLE:
+"That sounds like a content review workflow! Who typically reviews the content before it goes live?"
+
+BAD RESPONSE EXAMPLE:
+"**Great choice!** Here are some questions:
+- What type of content?
+- Who reviews it?
+- What happens after approval?"
+
+### Technical Reference (for YAML generation)
+
+Step types: `agent_task`, `human_approval`, `gateway`, `notification`, `sub_process`
+
+### YAML Schema Reference
+
+CRITICAL: Fields go directly in the step, NOT nested in a "config" block!
+
+```yaml
+name: my-process-name        # Required, kebab-case
+version: "1.0"               # Major.minor format
+description: "What this does"
+
+trigger:
+  type: manual               # manual | schedule | webhook
+
+steps:
+  - id: step-1
+    name: "First Step"
+    type: agent_task
+    agent: "my-agent"        # Required - directly in step
+    message: "Do something"  # Required - directly in step
+
+  - id: step-2
+    name: "Approval Gate"
+    type: human_approval
+    depends_on: [step-1]
+    approvers: ["admin@example.com"]  # Directly in step
+    timeout_hours: 24
+
+  - id: step-3
+    name: "Notify Team"
+    type: notification
+    depends_on: [step-2]
+    channel: email
+    recipients: ["team@example.com"]
+    message: "Process completed!"
+```
+
+### Workflow Patterns
+
+**Sequential (one after another):**
+```yaml
+steps:
+  - id: a
+  - id: b
+    depends_on: [a]
+  - id: c
+    depends_on: [b]
+```
+
+**Parallel (run simultaneously):**
+```yaml
+steps:
+  - id: a
+  - id: b    # No depends_on = runs in parallel with a
+  - id: c
+    depends_on: [a, b]  # Waits for both
+```
+
+**Approval workflow:**
+```yaml
+steps:
+  - id: prepare
+    type: agent_task
+  - id: review
+    type: human_approval
+    depends_on: [prepare]
+  - id: execute
+    type: agent_task
+    depends_on: [review]
+```
+
+### Best Practices for Generated YAML
+
+1. **Use descriptive names** - `analyze-customer-feedback` not `step1`
+2. **Add comments** - Help users understand each step
+3. **Use variable interpolation** - `{{input.fieldName}}` for inputs
+4. **Include error handling** - Set appropriate timeouts
+5. **Keep it simple** - Start minimal, user can expand later
+
+### Example Conversation
+
+**User:** "I want to automate content review before publishing"
+
+**You should:**
+1. Ask what content (blog, social, docs?)
+2. Ask who needs to approve
+3. Check what agents are available: `mcp__trinity__list_agents`
+4. Generate YAML with:
+   - Analysis step (agent_task)
+   - Review step (human_approval)
+   - Publish/notify step
+
+**Output format:**
+Always wrap generated YAML in \`\`\`yaml code blocks so the UI can detect and offer "Apply to Editor" functionality.
+
+### Using MCP Tools
+
+You can help users by checking available resources:
+
+```
+mcp__trinity__list_agents  # Show available agents with status
+mcp__trinity__get_agent    # Get details about a specific agent
+```
+
+When suggesting agents:
+- Note which are running vs stopped
+- Suggest starting stopped agents if needed
+- Match agent capabilities to task requirements

config/process-docs/editor-help.json

@@ -0,0 +1,233 @@
+{
+  "name": {
+    "title": "Process Name",
+    "description": "A unique identifier for your process. Used in URLs and API calls.",
+    "type": "string",
+    "required": true,
+    "example": "content-review-workflow",
+    "docs_link": "/processes/docs/reference/yaml-schema"
+  },
+  "version": {
+    "title": "Version",
+    "description": "Semantic version of your process definition. Increment when making changes.",
+    "type": "string",
+    "required": false,
+    "default": "1.0",
+    "example": "1.0.0",
+    "docs_link": "/processes/docs/reference/yaml-schema"
+  },
+  "description": {
+    "title": "Description",
+    "description": "Human-readable description of what this process does. Shown in the process list.",
+    "type": "string",
+    "required": false,
+    "example": "Automates content review with AI analysis and human approval"
+  },
+  "triggers": {
+    "title": "Triggers",
+    "description": "Define how this process can be started. Processes can have multiple triggers.",
+    "type": "array",
+    "required": true,
+    "example": "triggers:\n  - type: manual\n    id: manual-start",
+    "docs_link": "/processes/docs/reference/triggers"
+  },
+  "triggers.type": {
+    "title": "Trigger Type",
+    "description": "How the process is started: 'manual' (user-initiated), 'webhook' (HTTP POST), or 'schedule' (cron).",
+    "type": "string",
+    "required": true,
+    "options": ["manual", "webhook", "schedule"],
+    "docs_link": "/processes/docs/reference/triggers"
+  },
+  "triggers.id": {
+    "title": "Trigger ID",
+    "description": "Unique identifier for this trigger within the process.",
+    "type": "string",
+    "required": true,
+    "example": "daily-trigger"
+  },
+  "triggers.schedule": {
+    "title": "Schedule (Cron)",
+    "description": "Cron expression for scheduled triggers. Format: minute hour day-of-month month day-of-week.",
+    "type": "string",
+    "required": false,
+    "example": "0 9 * * MON-FRI",
+    "docs_link": "/processes/docs/reference/triggers"
+  },
+  "steps": {
+    "title": "Steps",
+    "description": "The sequence of tasks that make up your workflow. Steps can run sequentially or in parallel.",
+    "type": "array",
+    "required": true,
+    "docs_link": "/processes/docs/getting-started/step-types"
+  },
+  "steps.id": {
+    "title": "Step ID",
+    "description": "Unique identifier for this step. Referenced by depends_on and variable interpolation.",
+    "type": "string",
+    "required": true,
+    "example": "analyze-content"
+  },
+  "steps.name": {
+    "title": "Step Name",
+    "description": "Human-readable name shown in the execution timeline.",
+    "type": "string",
+    "required": false,
+    "example": "Analyze Content Quality"
+  },
+  "steps.type": {
+    "title": "Step Type",
+    "description": "The type of step determines what action is performed.",
+    "type": "string",
+    "required": true,
+    "options": ["agent_task", "human_approval", "gateway", "timer", "notification", "sub_process"],
+    "docs_link": "/processes/docs/getting-started/step-types"
+  },
+  "steps.agent": {
+    "title": "Agent",
+    "description": "The agent (by name or ID) that will execute this task. Must be a running agent.",
+    "type": "string",
+    "required": true,
+    "example": "content-reviewer"
+  },
+  "steps.message": {
+    "title": "Message/Prompt",
+    "description": "The instruction sent to the agent. Supports variable interpolation with {{...}} syntax.",
+    "type": "string",
+    "required": true,
+    "example": "Analyze this content: {{input.content}}",
+    "docs_link": "/processes/docs/reference/variables"
+  },
+  "steps.depends_on": {
+    "title": "Dependencies",
+    "description": "List of step IDs that must complete before this step runs. Creates sequential/parallel flow.",
+    "type": "array",
+    "required": false,
+    "example": "depends_on:\n  - step-1\n  - step-2",
+    "docs_link": "/processes/docs/patterns/sequential"
+  },
+  "steps.timeout": {
+    "title": "Timeout",
+    "description": "Maximum time this step can run. Format: number + unit (s, m, h). Step fails if exceeded.",
+    "type": "string",
+    "required": false,
+    "default": "30m",
+    "example": "10m"
+  },
+  "steps.output_key": {
+    "title": "Output Key",
+    "description": "Name for storing this step's output. Access later with {{steps.step_id.output.key}}.",
+    "type": "string",
+    "required": false,
+    "example": "analysis_result"
+  },
+  "steps.title": {
+    "title": "Approval Title",
+    "description": "Title shown in the approval dialog for human_approval steps.",
+    "type": "string",
+    "required": false,
+    "example": "Content Review Required"
+  },
+  "steps.description": {
+    "title": "Approval Description",
+    "description": "Detailed description shown to approvers. Supports variable interpolation.",
+    "type": "string",
+    "required": false
+  },
+  "steps.assignees": {
+    "title": "Assignees",
+    "description": "List of users who can approve this step. If empty, any user can approve.",
+    "type": "array",
+    "required": false,
+    "example": "assignees:\n  - alice\n  - bob"
+  },
+  "steps.gateway_type": {
+    "title": "Gateway Type",
+    "description": "How routing decisions are made: 'exclusive' (first match), 'parallel' (all routes).",
+    "type": "string",
+    "required": false,
+    "default": "exclusive",
+    "options": ["exclusive", "parallel"],
+    "docs_link": "/processes/docs/patterns/parallel"
+  },
+  "steps.routes": {
+    "title": "Gateway Routes",
+    "description": "Conditional routing rules for gateway steps. Each route has a condition and target.",
+    "type": "array",
+    "required": false,
+    "docs_link": "/processes/docs/getting-started/step-types"
+  },
+  "steps.routes.condition": {
+    "title": "Route Condition",
+    "description": "Expression that determines if this route is taken. Uses {{...}} variable syntax.",
+    "type": "string",
+    "required": true,
+    "example": "{{steps.analyze.output.score}} >= 80"
+  },
+  "steps.routes.target": {
+    "title": "Route Target",
+    "description": "Step ID to execute if this route's condition is true.",
+    "type": "string",
+    "required": true
+  },
+  "steps.default_route": {
+    "title": "Default Route",
+    "description": "Step ID to execute if no route conditions match.",
+    "type": "string",
+    "required": false
+  },
+  "steps.delay": {
+    "title": "Timer Delay",
+    "description": "How long to wait for timer steps. Format: number + unit (s, m, h, d).",
+    "type": "string",
+    "required": true,
+    "example": "30m"
+  },
+  "steps.process_id": {
+    "title": "Sub-Process ID",
+    "description": "The ID of another process to execute as a sub-process.",
+    "type": "string",
+    "required": true,
+    "docs_link": "/processes/docs/getting-started/step-types"
+  },
+  "steps.channel": {
+    "title": "Notification Channel",
+    "description": "Where to send notifications: 'email', 'slack', or 'webhook'.",
+    "type": "string",
+    "required": true,
+    "options": ["email", "slack", "webhook"]
+  },
+  "steps.recipients": {
+    "title": "Recipients",
+    "description": "List of recipients for the notification. Format depends on channel type.",
+    "type": "array",
+    "required": false,
+    "example": "recipients:\n  - user@example.com"
+  },
+  "outputs": {
+    "title": "Outputs",
+    "description": "Define process outputs that can be accessed after completion.",
+    "type": "array",
+    "required": false,
+    "docs_link": "/processes/docs/reference/yaml-schema"
+  },
+  "outputs.name": {
+    "title": "Output Name",
+    "description": "Name for this output value.",
+    "type": "string",
+    "required": true
+  },
+  "outputs.value": {
+    "title": "Output Value",
+    "description": "Expression to compute the output value. Uses {{...}} variable syntax.",
+    "type": "string",
+    "required": true,
+    "example": "{{steps.final-step.output}}"
+  },
+  "default": {
+    "title": "YAML Process Definition",
+    "description": "Define automated workflows using YAML. Processes orchestrate AI agents through steps with dependencies, conditions, and approvals.",
+    "type": "object",
+    "docs_link": "/processes/docs/getting-started/what-are-processes"
+  }
+}