skills: add deploy-cluster and destroy-cluster AI skills

.agents/skills/README.md

@@ -0,0 +1,13 @@
+# Claude Skills
+
+AI-powered skills for the bootc-operator project. These are reusable instructions that AI agents (Claude Code, Copilot, etc.) can invoke to perform common development tasks — so you don't have to remember every command or worry about missing a step.
+
+Each skill lives in its own directory as a `SKILL.md` file. When invoked, the agent follows the instructions step by step, checking prerequisites, running commands, and verifying results on your behalf.
+
+## Available Skills
+
+| Skill | Description |
+|-------|-------------|
+| [deploy-cluster](deploy-cluster/SKILL.md) | Build and deploy bootc-operator to a local bink cluster |
+| [destroy-cluster](destroy-cluster/SKILL.md) | Tear down a local bink cluster and clean up resources |
+| [k8s-api-review](k8s-api-review/SKILL.md) | Review Kubernetes API changes for convention compliance |

.agents/skills/deploy-cluster/SKILL.md

@@ -0,0 +1,238 @@
+---
+name: deploy-cluster
+description: Deploy the bootc-operator to a local bink cluster for development and testing
+---
+
+## What I do
+
+I guide you through building and deploying the bootc-operator to a local bink cluster from scratch. I check all prerequisites, build the operator image, create the cluster, deploy the operator, and verify everything is healthy.
+
+## When to use me
+
+- Setting up a local development cluster for the first time
+- Redeploying after code changes
+- Troubleshooting a broken cluster by tearing down and redeploying
+- Preparing to run e2e tests
+
+---
+
+## Instructions for AI Agent
+
+All commands run from the project root: `/home/ptalgulk/VSCode/bootc/bootc-operator`
+
+### Step 1: Check prerequisites
+
+Run each check. If any tool is missing, tell the user how to install it before continuing.
+
+```bash
+echo "=== Checking prerequisites ==="
+
+echo -n "Go: "; go version 2>/dev/null || echo "MISSING - install from https://go.dev/dl/ (need >= 1.26)"
+echo -n "Make: "; make --version 2>/dev/null | head -1 || echo "MISSING - dnf install make"
+echo -n "Podman: "; podman --version 2>/dev/null || echo "MISSING - dnf install podman"
+echo -n "kubectl: "; kubectl version --client 2>/dev/null | head -1 || echo "MISSING - https://kubernetes.io/docs/tasks/tools/"
+echo -n "bink: "; bink version 2>/dev/null || echo "MISSING - go install github.com/bootc-dev/bink@latest"
+echo -n "skopeo: "; skopeo --version 2>/dev/null || echo "MISSING - dnf install skopeo"
+```
+
+Check the Go version meets the `go.mod` requirement (>= 1.26):
+
+```bash
+go_ver=$(go version | grep -oP 'go\K[0-9]+\.[0-9]+')
+echo "Go version: $go_ver (need >= 1.26)"
+```
+
+Check Podman socket is running:
+
+```bash
+systemctl --user is-active podman.socket
+```
+
+If inactive, start it:
+
+```bash
+systemctl --user start podman.socket
+```
+
+Check port 5000 is available (used by the local registry):
+
+```bash
+ss -tlnp | grep :5000 || echo "Port 5000 is available"
+```
+
+If port 5000 is already in use by something other than a bink registry, tell the user to free it.
+
+**Stop here if any prerequisite is missing. Do not proceed until all tools are installed.**
+
+### Step 2: Build the operator image
+
+```bash
+make buildimg
+```
+
+**What this does:** Builds the `bootc-operator:dev` container image using Podman. The image contains both the controller and daemon binaries.
+
+**Expected output:** The last line should say `Successfully tagged localhost/bootc-operator:dev`.
+
+If the build fails, check:
+- Is `podman.socket` running? (`systemctl --user start podman.socket`)
+- Are there Go compilation errors? (fix the code first)
+
+### Step 3: Deploy to bink cluster
+
+```bash
+make deploy-bink
+```
+
+**What this does (in order):**
+1. Starts a local container registry at `localhost:5000`
+2. Pulls the bootc node disk image and pushes it to the registry
+3. Creates a bink cluster named `e2e` with a control-plane node (skips if already running)
+4. Builds a derived node image for update testing
+5. Pushes the operator image to the in-cluster registry
+6. Deploys the operator via kustomize with `--allow-insecure-registry` and short tag resolution
+7. Waits for the controller deployment to roll out
+
+**Expected output:** The last line should say `deployment "bootc-operator-controller-manager" successfully rolled out`.
+
+This step takes 2-5 minutes on first run (pulling images). Subsequent runs are faster.
+
+To use a separate dev cluster (so e2e tests don't interfere):
+
+```bash
+make deploy-bink BINK_CLUSTER_NAME=dev
+```
+
+### Step 4: Set up kubeconfig
+
+Tell the user to run this in their terminal:
+
+```bash
+export KUBECONFIG=$(pwd)/kubeconfig-e2e
+```
+
+For a named cluster (e.g. `dev`):
+
+```bash
+export KUBECONFIG=$(pwd)/kubeconfig-dev
+```
+
+### Step 5: Verify deployment health
+
+Run these checks and report results to the user:
+
+```bash
+echo "=== Cluster Health Check ==="
+
+echo "--- Nodes ---"
+kubectl get nodes
+
+echo "--- CRDs ---"
+kubectl get crd | grep bootc
+
+echo "--- Operator Pods ---"
+kubectl -n bootc-operator get pods
+
+echo "--- Controller Deployment ---"
+kubectl -n bootc-operator get deployment bootc-operator-controller-manager
+
+echo "--- Daemon DaemonSet ---"
+kubectl -n bootc-operator get daemonset bootc-operator-daemon
+```
+
+**Healthy state checklist:**
+- [ ] Controller node is `Ready`
+- [ ] CRDs `bootcnodes.node.bootc.dev` and `bootcnodepools.node.bootc.dev` exist
+- [ ] Controller manager pod is `Running` (1/1)
+- [ ] Daemon DaemonSet has desired count matching available nodes
+
+If any check fails:
+
+**Controller pod not running:**
+```bash
+kubectl -n bootc-operator describe deployment bootc-operator-controller-manager
+kubectl -n bootc-operator logs deployment/bootc-operator-controller-manager
+```
+
+**Daemon pod not running:**
+```bash
+kubectl -n bootc-operator describe daemonset bootc-operator-daemon
+kubectl -n bootc-operator get events -n bootc-operator --sort-by='.lastTimestamp' | tail -20
+```
+
+**CRDs missing:**
+```bash
+make install KUBECONFIG=./kubeconfig-e2e
+```
+
+**Node not Ready:**
+```bash
+kubectl describe node controller
+```
+
+### Step 6 (optional): Smoke test
+
+If the user wants to verify end-to-end functionality:
+
+```bash
+make e2e V=1 RUN=TestControllerMembership
+```
+
+This runs a single fast test that provisions a worker node, creates a pool, and verifies a BootcNode is created. Takes about 40 seconds.
+
+To run the full e2e suite:
+
+```bash
+make e2e V=1
+```
+
+To run a specific test:
+
+```bash
+make e2e V=1 RUN=TestUpdateReboot
+```
+
+### Redeploying after code changes
+
+```bash
+make buildimg
+make deploy-bink
+```
+
+If API types in `api/` were changed, regenerate manifests first:
+
+```bash
+make manifests generate
+make buildimg
+make deploy-bink
+```
+
+### Gathering diagnostic logs
+
+```bash
+make gather-bink
+```
+
+Logs are written to `_output/logs/gather-bink/`.
+
+### Troubleshooting
+
+| Symptom | Fix |
+|---------|-----|
+| `podman.socket` not running | `systemctl --user start podman.socket` |
+| Port 5000 in use | Stop the process using it, or tear down and recreate |
+| Cluster unhealthy | `make teardown-bink && make buildimg && make deploy-bink` |
+| Image not updating | Verify `make buildimg` succeeded, check `skopeo inspect --tls-verify=false docker://localhost:5000/bootc-operator-e2e:latest` |
+| CRDs not installed | `make install KUBECONFIG=./kubeconfig-e2e` |
+| `bink` not found | `go install github.com/bootc-dev/bink@latest` |
+| `skopeo` not found | `dnf install skopeo` |
+| Go version too old | Install Go >= 1.26 from https://go.dev/dl/ |
+
+### Key variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `IMG` | `bootc-operator:dev` | Operator container image name |
+| `CONTAINER_TOOL` | `podman` | Container runtime |
+| `BINK_CLUSTER_NAME` | `e2e` | Cluster name (change to avoid conflicts) |
+| `KUBECONFIG_BINK` | `./kubeconfig-e2e` | Kubeconfig path (derived from cluster name) |

.agents/skills/destroy-cluster/SKILL.md

@@ -0,0 +1,115 @@
+---
+name: destroy-cluster
+description: Tear down a local bink cluster and clean up all associated resources
+---
+
+## What I do
+
+I safely tear down a local bink cluster used for bootc-operator development and testing. I verify the cluster exists, remove it, clean up the kubeconfig, and confirm everything is gone.
+
+## When to use me
+
+- Done with development/testing and want to free resources
+- Cluster is in a bad state and you want a clean start (destroy then redeploy)
+- Cleaning up after e2e test runs
+
+---
+
+## Instructions for AI Agent
+
+All commands run from the project root: `/home/ptalgulk/VSCode/bootc/bootc-operator`
+
+### Step 1: Check prerequisites
+
+```bash
+echo -n "Make: "; make --version 2>/dev/null | head -1 || echo "MISSING"
+echo -n "bink: "; bink version 2>/dev/null || echo "MISSING"
+echo -n "Podman: "; podman --version 2>/dev/null || echo "MISSING"
+```
+
+### Step 2: Check if cluster exists
+
+```bash
+bink cluster list 2>/dev/null
+```
+
+If the default `e2e` cluster is not listed, tell the user there is nothing to tear down.
+
+For a named cluster:
+
+```bash
+bink cluster list 2>/dev/null | grep <cluster-name>
+```
+
+### Step 3: Tear down the cluster
+
+Default cluster (`e2e`):
+
+```bash
+make teardown-bink
+```
+
+Named cluster:
+
+```bash
+make teardown-bink BINK_CLUSTER_NAME=<name>
+```
+
+**What this does:**
+1. Removes all bink worker nodes
+2. Stops and removes the control-plane container
+3. Stops the local container registry
+
+This typically takes 10-30 seconds.
+
+### Step 4: Clean up kubeconfig
+
+```bash
+rm -f kubeconfig-e2e
+```
+
+For a named cluster:
+
+```bash
+rm -f kubeconfig-<name>
+```
+
+Tell the user to unset KUBECONFIG in their terminal if it was pointing at the deleted file:
+
+```bash
+unset KUBECONFIG
+```
+
+### Step 5: Verify teardown
+
+```bash
+echo "=== Teardown Verification ==="
+
+echo "--- Remaining bink clusters ---"
+bink cluster list 2>/dev/null || echo "No clusters found"
+
+echo "--- Remaining containers (bootc/bink related) ---"
+podman ps -a --filter "name=e2e" --format "{{.Names}}" 2>/dev/null || echo "None"
+
+echo "--- Registry on port 5000 ---"
+ss -tlnp | grep :5000 || echo "Port 5000 is free"
+
+echo "--- Kubeconfig file ---"
+ls kubeconfig-e2e 2>/dev/null || echo "Kubeconfig removed"
+```
+
+**Clean state checklist:**
+- [ ] No bink clusters listed (or the target cluster is gone)
+- [ ] No containers with the cluster name running
+- [ ] Port 5000 is free (registry stopped)
+- [ ] Kubeconfig file removed
+
+### Troubleshooting
+
+| Symptom | Fix |
+|---------|-----|
+| `make teardown-bink` fails | Try `bink cluster delete e2e --force` directly |
+| Containers still running | `podman rm -f <container-name>` |
+| Port 5000 still in use | `podman stop <registry-container>` then `podman rm <registry-container>` |
+| Volumes left behind | `podman volume prune` (ask user first — affects all unused volumes) |
+| Teardown hangs | Ctrl+C, then manually remove with `bink cluster delete e2e --force` |

.claude/skills/README.md

@@ -0,0 +1,13 @@
+# Claude Skills
+
+AI-powered skills for the bootc-operator project. These are reusable instructions that AI agents (Claude Code, Copilot, etc.) can invoke to perform common development tasks — so you don't have to remember every command or worry about missing a step.
+
+Each skill lives in its own directory as a `SKILL.md` file. When invoked, the agent follows the instructions step by step, checking prerequisites, running commands, and verifying results on your behalf.
+
+## Available Skills
+
+| Skill | Description |
+|-------|-------------|
+| [deploy-cluster](deploy-cluster/SKILL.md) | Build and deploy bootc-operator to a local bink cluster |
+| [destroy-cluster](destroy-cluster/SKILL.md) | Tear down a local bink cluster and clean up resources |
+| [k8s-api-review](k8s-api-review/SKILL.md) | Review Kubernetes API changes for convention compliance |