Workflow integration

README.md

@@ -90,7 +90,56 @@ Set workspace hosts and Unity Catalog grants per `docs/setup.md` in the rendered
 
 If the bundle was scaffolded inside a Databricks **Git folder** in the workspace (the recommended path for Genie Code users), the workspace UI also surfaces a **Deployments panel** on the bundle that lets you pick a target and deploy with one click — no terminal required. This matches the layout produced by the workspace UI's native "Create → Bundle" flow.
 
-> **Coming next:** production lifecycle patterns (evaluation gates, governance posture, monitoring, feedback loops) are in active design — they'll ship as Innovate-driven workflows on top of the scaffold. Subscribe to releases for updates.
+---
+
+## Production lifecycle
+
+The scaffold generates the envelope. The **AgentOps Lifecycle** skill guides
+the project through the complete Single-Account Single-Agent lifecycle — from
+first data ingestion to production monitoring — across three phases and 10 steps:
+
+| Phase | Steps | What happens |
+|-------|-------|--------------|
+| **Dev** | 1–5 | Data prep + Vector Search indexing, agent implementation with MLflow tracing, offline eval gate, SME calibration |
+| **Staging** | 6–7 | CI gate (unit tests + bundle validate + eval gate on PR), staging deploy + integration tests |
+| **Production** | 8–10 | CD deploy, batch inferencing baseline, real-time monitoring + SME feedback loop |
+
+MLflow is the operational spine at every level: experiment tracking in dev,
+eval gate in CI, trace logging in prod, user feedback via `mlflow.log_feedback()`.
+The `evaluation/gate.py` pattern (generated by the scaffold) runs at three
+points — locally in dev, in CI on every PR, and against production data before
+users are admitted — so quality regressions are caught before they reach users.
+
+### Using the lifecycle skill
+
+Install the plugin, then ask your coding assistant to continue from where the
+scaffold left off:
+
+```bash
+# Install (Claude Code / Cursor)
+./plugin/skills/install_skills.sh
+
+# Then in the assistant:
+/agentops-lifecycle
+# or: "walk me through the agentops lifecycle"
+```
+
+In **Genie Code**, run `install_genie_code_skills.py` as a notebook to install,
+then say "walk me through the agentops lifecycle."
+
+The skill provides step-by-step agent actions, copy-paste code examples
+grounded in the generated project files, validation criteria, and common issue
+resolutions for each of the 10 steps.
+
+### Machine-readable workflow definition
+
+`workflows/single-account-single-agent.json` is a structured definition of the
+10-step lifecycle with all actions, validations, escalation hints, and phase
+metadata. It can be consumed programmatically by workflow engines (e.g.,
+`/innovate`) or used as a reference when building custom automation on top of
+the scaffold.
+
+---
 
 ## Documentation
 

plugin/README.md

@@ -4,15 +4,27 @@ A portable plugin that scaffolds AgentOps Stacks projects from inside a coding a
 
 The plugin and the [DAB template](../template/) share the same scaffold contract (`.agentops-stacks/manifest.yml`). The CLI alone is sufficient for scaffolding; the plugin is the additive on-ramp for users working inside a coding assistant.
 
-## What this plugin ships today
+## What this plugin ships
 
-- One skill: **`agentops-stacks`** — scaffolds a new project (DAB layout, dev/staging/prod targets, UC schema and volume, MLflow experiment, CI/CD wiring for one of GitHub Actions, GitHub Actions for GHES, GitLab, or Azure DevOps).
-- One command: **`/init-agentops-stacks`** — discoverability wrapper around the skill for Claude Code / Cursor users.
-- Installers under `skills/` that mirror [ai-dev-kit's pattern](https://github.com/databricks-solutions/ai-dev-kit#installation):
-  - `install_skills.sh` — CLI installer. Lands the skill into `.claude/skills/agentops-stacks/` locally, with optional `--install-to-genie` to also upload to the workspace.
-  - `install_genie_code_skills.py` — Databricks notebook for in-workspace installs without a local clone.
+### Skills
 
-Production patterns (evaluation gates, governance posture, monitoring, feedback loops) are not in this plugin yet. They land as separate skills as the project matures.
+| Skill | Trigger | What it does |
+|-------|---------|--------------|
+| **`agentops-stacks`** | `scaffold a new agentops project` | Scaffolds a new project (DAB layout, dev/staging/prod targets, UC schema and volume, MLflow experiment, CI/CD wiring). One-time use at project start. |
+| **`agentops-lifecycle`** | `walk me through the agentops lifecycle` | Guides an existing scaffold through the complete Single-Account Single-Agent lifecycle — data prep, agent dev, eval gate, SME calibration, CI/CD promotion, batch eval baseline, and production monitoring. 10 steps across dev → staging → prod. |
+
+### Commands
+
+| Command | Skill |
+|---------|-------|
+| `/init-agentops-stacks` | `agentops-stacks` |
+| `/agentops-lifecycle` | `agentops-lifecycle` |
+
+### Installers
+
+Under `skills/`, mirroring [ai-dev-kit's pattern](https://github.com/databricks-solutions/ai-dev-kit#installation):
+- `install_skills.sh` — CLI installer. Lands skills into `.claude/skills/` locally, with optional `--install-to-genie` to also upload to the workspace.
+- `install_genie_code_skills.py` — Databricks notebook for in-workspace installs without a local clone.
 
 ## Prerequisites
 
@@ -82,15 +94,21 @@ Prerequisites: `databricks` CLI installed, a coding assistant that loads `.claud
 plugin/
 ├── README.md                                    # this file
 ├── commands/
-│   └── init-agentops-stacks.md                  # CC/Cursor slash command
+│   ├── init-agentops-stacks.md                  # scaffold command
+│   └── agentops-lifecycle.md                    # lifecycle command
 └── skills/
     ├── install_skills.sh                        # local + Genie upload installer
     ├── install_genie_code_skills.py             # in-workspace notebook installer
-    └── agentops-stacks/
-        └── SKILL.md                             # skill instructions for the assistant
+    ├── agentops-stacks/
+    │   └── SKILL.md                             # scaffold skill
+    └── agentops-lifecycle/
+        └── SKILL.md                             # lifecycle skill (10-step dev→prod guide)
 ```
 
-The skill is a single `SKILL.md`. There's no Python renderer and no vendored template tree — `databricks bundle init` reads the template directly from this repo at scaffold time.
+Each skill is a single `SKILL.md`. There's no Python renderer and no vendored
+template tree — `databricks bundle init` reads the template directly from this
+repo at scaffold time, and the lifecycle skill references generated project
+files by their scaffold-relative paths.
 
 ## Known UX notes
 

plugin/commands/agentops-lifecycle.md

@@ -0,0 +1,22 @@
+---
+description: >
+  Guide an agentops-stacks project through its complete production lifecycle —
+  data prep, agent dev, eval gates, CI/CD promotion, and production monitoring.
+  Run after `databricks bundle init` (or `/init-agentops-stacks`).
+---
+
+Use the `agentops-lifecycle` skill.
+
+This command guides an existing agentops-stacks scaffold through the
+Single-Account Single-Agent lifecycle:
+
+- Steps 1–5 (Dev): data preparation, agent implementation, offline evaluation,
+  SME calibration
+- Steps 6–7 (Staging): CI gate, integration tests
+- Steps 8–10 (Production): CD deploy, batch inferencing baseline, monitoring
+
+**Prerequisite:** `.agentops-stacks/manifest.yml` must exist. Run
+`/init-agentops-stacks` first if you haven't scaffolded yet.
+
+Defer to the skill's SKILL.md for full step-by-step guidance, code examples,
+validation criteria, and common issue resolutions.