m9m-system-orchestrator

name	m9m-system-orchestrator
description	End-to-end control plane for m9m. Use when a request spans discovery, workflow creation, execution, debugging, plugin/node extension, and lifecycle management.

m9m System Orchestrator

Use this skill when the user asks for outcomes that require multiple m9m surfaces, for example:

• "Build and run an automation pipeline, then harden it."
• "Set up agent workflows and keep them reliable in production."
• "Manage workflows, nodes, and execution state across the whole system."

Control Loop

1. Discover current capabilities and constraints first.
2. Design or update workflow topology and node contracts.
3. Execute with sync or async mode based on expected duration.
4. Observe status, logs, and node outputs.
5. Recover via retry/cancel/fix.
6. Promote reusable assets (workflows/plugins), then govern lifecycle.

Skill Routing

• Use m9m-capability-discovery for node/workflow/plugin inventory.
• Use m9m-workflow-factory for workflow create/update/validate/activate.
• Use m9m-execution-ops for execution run/wait/cancel/retry.
• Use m9m-debug-reliability for failures and performance regressions.
• Use m9m-plugin-node-builder for custom JavaScript/REST nodes.
• Use m9m-cli-agent-orchestrator for CLI agent execution with sandboxing.
• Use m9m-reuse-lifecycle for duplication, tagging, rollout, and cleanup.

MCP-First, REST-Fallback

• Prefer MCP tools when connected to mcp-server.
• If MCP is unavailable, use REST endpoints in documentation/docs/api/*.md.
• Keep the same lifecycle semantics between MCP and REST.

Guardrails

• Validate before create/update (workflow_validate).
• Track workflowId and executionId explicitly in outputs.
• Treat unauthenticated API access as internal/experimental only.
• Treat Kubernetes deployment files as experimental; prefer single-binary operations for official paths.