fix(vm): remove unsupported hibernate action

README.md

@@ -159,7 +159,7 @@ Credentials are shared with the CLI — run `verda auth login` first.
 
 ### Safety contract
 
-Tools that create billed resources (`create_vm`, `create_volume`) require `confirm: true`, and `vm_action` requires it for the destructive actions (`shutdown`, `force_shutdown`, `hibernate`, `delete`) — the same gate as `--yes` in `--agent` mode. Without it the tool fails with `CONFIRMATION_REQUIRED` in the [agent error format](docs/agent-errors.md#mcp-server-tools), so the agent can present the exact action to the user and retry after approval.
+Tools that create billed resources (`create_vm`, `create_volume`) require `confirm: true`, and `vm_action` requires it for the destructive actions (`shutdown`, `force_shutdown`, `delete`) — the same gate as `--yes` in `--agent` mode. Without it the tool fails with `CONFIRMATION_REQUIRED` in the [agent error format](docs/agent-errors.md#mcp-server-tools), so the agent can present the exact action to the user and retry after approval.
 
 `vm_action` reports `status: "accepted"` once the API accepts the action; only `wait: true` polls the instance to its expected status and reports `"completed"` (`create_vm` waits by default). Argument types are validated strictly — wrong JSON types and unknown enum values fail with `VALIDATION_ERROR` instead of being silently coerced. The full tool/parameter reference lives in [cmd/mcp/README.md](internal/verda-cli/cmd/mcp/README.md).
 

docs/ai-skills.md

@@ -82,7 +82,6 @@ verda template edit my-tmpl  # Edit an existing template
 List my running Verda Cloud VMs
 Shut down my Verda Cloud training VM
 Start my Verda Cloud gpu-runner instance back up
-Hibernate my Verda Cloud dev box
 Delete the Verda Cloud instance I'm not using anymore
 ```
 

docs/commands.md

@@ -9,7 +9,7 @@ Full reference for all Verda CLI commands and flags.
 | `verda vm create` | Create a VM (interactive wizard or flags) |
 | `verda vm list` | List and inspect VM instances |
 | `verda vm describe` | Show detailed info about a single VM |
-| `verda vm action` | Start, shutdown, hibernate, or delete a VM |
+| `verda vm action` | Start, shutdown, or delete a VM |
 | `verda vm availability` | Show available instance types with pricing |
 
 ## SSH

internal/skills/files/verda-reference.md

@@ -16,7 +16,6 @@ All commands: `--agent -o json` (except `verda ssh` and `verda auth show`).
 | "VM info", "instance info", "describe", "show VM" | `vm describe <id>` |
 | "start", "boot", "power on" | `vm start <id>` |
 | "stop", "shut down", "power off" | `vm shutdown <id>` (alias: `stop`) |
-| "hibernate", "suspend", "sleep" | `vm hibernate <id>` |
 | "delete VM", "delete instance", "remove", "destroy", "terminate" | `vm delete <id>` (alias: `rm`) |
 | "template", "saved config", "preset", "my templates" | `template list` (alias: `tmpl`) |
 | "deploy from template", "use template", "quick deploy" | `vm create --from <name>` |
@@ -73,7 +72,6 @@ All commands: `--agent -o json` (except `verda ssh` and `verda auth show`).
 | `verda vm describe <id> -o json` | — |
 | `verda vm start <id> --wait` | `--yes` in agent mode |
 | `verda vm shutdown <id> --wait` | `--yes` in agent mode. Alias: `stop` |
-| `verda vm hibernate <id> --wait` | `--yes` in agent mode |
 | `verda vm delete <id> --wait` | `--yes` **required**. `--with-volumes` to also delete attached volumes. Alias: `rm` |
 
 Batch operations: `--all` with `--status` and/or `--hostname` (glob pattern) to target multiple VMs.

internal/verda-cli/cmd/mcp/CLAUDE.md

@@ -15,7 +15,7 @@
 
 ### Confirm gates (hard requirement, mirrors `--yes`)
 - Billing: `create_vm`, `create_volume` — `confirm: true` always required (schema marks it Required, but mcp-go does not enforce schemas — the handler gate is the enforcement)
-- Destructive `vm_action`: `shutdown`, `force_shutdown`, `hibernate`, `delete` — gated via the `vmActions` table's `destructive` flag
+- Destructive `vm_action`: `shutdown`, `force_shutdown`, `delete` — gated via the `vmActions` table's `destructive` flag
 - The gate runs BEFORE `verdaClient()`/API calls; a refused call must have zero side effects
 - Agent-facing codes: `CONFIRMATION_REQUIRED`, `VALIDATION_ERROR`, `MISSING_REQUIRED_FLAGS` — same names as the CLI `--agent` contract
 

internal/verda-cli/cmd/mcp/README.md

@@ -40,7 +40,7 @@ Credentials are shared with the CLI — run `verda auth login` first in the same
 
 ## Confirmation contract
 
-Tools that create billed resources (`create_vm`, `create_volume`) or perform destructive actions (`vm_action` with `shutdown`, `force_shutdown`, `hibernate`, `delete`) require the boolean argument `confirm: true`. This mirrors `--yes` in the CLI's `--agent` mode.
+Tools that create billed resources (`create_vm`, `create_volume`) or perform destructive actions (`vm_action` with `shutdown`, `force_shutdown`, `delete`) require the boolean argument `confirm: true`. This mirrors `--yes` in the CLI's `--agent` mode.
 
 Without it, the tool fails with `isError: true` and a JSON payload following the CLI agent-error envelope (see [docs/agent-errors.md](../../../docs/agent-errors.md)):
 
@@ -54,7 +54,7 @@ Without it, the tool fails with `isError: true` and a JSON payload following the
 
 `vm_action` reports `status: "accepted"` by default — the API has accepted the action, nothing more. `shutdown` of a running VM is *not* done at that point; do not tell the user billing stopped.
 
-With `wait: true`, the tool polls the instance (up to 5 minutes) until it reaches the action's expected status (`start`→running, `shutdown`/`force_shutdown`/`hibernate`→offline) and then reports `status: "completed"` plus `instance_status`. A failed transition (instance enters `error`) is a tool error, not a success. `delete` is not polled and always returns `accepted`.
+With `wait: true`, the tool polls the instance (up to 5 minutes) until it reaches the action's expected status (`start`→running, `shutdown`/`force_shutdown`→offline) and then reports `status: "completed"` plus `instance_status`. A failed transition (instance enters `error`) is a tool error, not a success. `delete` is not polled and always returns `accepted`.
 
 `create_vm` keeps `wait: true` as its default and blocks until the instance is `running`; on poll failure it returns the instance plus `poll_error`/`poll_timed_out` fields.
 
@@ -64,7 +64,7 @@ mcp-go does not validate arguments against the declared schema. All handlers typ
 
 - numbers must be numbers — `"500"` for `os_volume_size_gb` is rejected (was: coerced to 0, silently applying the 50 GB default)
 - arrays must be arrays — `ssh_key_ids` as a bare string is rejected (was: dropped, attaching *all* account SSH keys)
-- enums are checked against their allowed sets — `storage_type` accepts only `NVMe`/`HDD`; `vm_action` `action` accepts only the five documented verbs; unknown `estimate_cost` `storage_type` errors listing the types from the API catalog instead of pricing storage at $0
+- enums are checked against their allowed sets — `storage_type` accepts only `NVMe`/`HDD`; `vm_action` `action` accepts only the four documented verbs; unknown `estimate_cost` `storage_type` errors listing the types from the API catalog instead of pricing storage at $0
 - missing required arguments produce `MISSING_REQUIRED_FLAGS`
 
 ## Files

internal/verda-cli/cmd/mcp/server.go

@@ -59,7 +59,7 @@ func NewLazyServer(getClient clientFunc) *Server {
 const serverInstructions = `Verda Cloud: GPU/CPU instances, volumes, SSH keys, object storage.
 
 CONFIRM GATE — tools that create billing or destructive changes (create_vm,
-create_volume, and vm_action with shutdown/force_shutdown/hibernate/delete)
+create_volume, and vm_action with shutdown/force_shutdown/delete)
 refuse to run unless you pass confirm: true. Show the user the exact target and
 its cost first, then retry with confirm. A refused call has no side effects.
 

internal/verda-cli/cmd/mcp/tools_test.go

@@ -157,7 +157,7 @@ func TestCreateVMConfirmGate(t *testing.T) {
 func TestVMActionConfirmGate(t *testing.T) {
 	t.Parallel()
 
-	destructive := []string{"shutdown", "force_shutdown", "hibernate", "delete"}
+	destructive := []string{"shutdown", "force_shutdown", "delete"}
 	for _, action := range destructive {
 		t.Run(action, func(t *testing.T) {
 			t.Parallel()

internal/verda-cli/cmd/mcp/tools_vm.go

@@ -81,10 +81,10 @@ func (s *Server) registerVMTools() {
 
 	s.mcpServer.AddTool(
 		mcp.NewTool("vm_action",
-			mcp.WithDescription("Perform an action on a VM: start, shutdown, force_shutdown, hibernate, or delete. Destructive actions (shutdown, force_shutdown, hibernate, delete) REQUIRE confirm: true — confirm with the user first; without it the tool fails with CONFIRMATION_REQUIRED. Returns status 'accepted' once the API has accepted the action; pass wait: true to poll until the instance reaches its expected status and report status 'completed' (a failed transition, e.g. the instance entering 'error', is a tool error). delete is not polled and always returns 'accepted'."),
+			mcp.WithDescription("Perform an action on a VM: start, shutdown, force_shutdown, or delete. Destructive actions (shutdown, force_shutdown, delete) REQUIRE confirm: true — confirm with the user first; without it the tool fails with CONFIRMATION_REQUIRED. Returns status 'accepted' once the API has accepted the action; pass wait: true to poll until the instance reaches its expected status and report status 'completed' (a failed transition, e.g. the instance entering 'error', is a tool error). delete is not polled and always returns 'accepted'."),
 			mcp.WithString("id", mcp.Required(), mcp.Description("Instance ID")),
-			mcp.WithString("action", mcp.Required(), mcp.Description("Action: start, shutdown, force_shutdown, hibernate, delete")),
-			mcp.WithBoolean("confirm", mcp.Description("Set true to confirm destructive actions (shutdown, force_shutdown, hibernate, delete). Not needed for start.")),
+			mcp.WithString("action", mcp.Required(), mcp.Description("Action: start, shutdown, force_shutdown, delete")),
+			mcp.WithBoolean("confirm", mcp.Description("Set true to confirm destructive actions (shutdown, force_shutdown, delete). Not needed for start.")),
 			mcp.WithBoolean("wait", mcp.Description("Poll until the instance reaches the action's expected status (default false: return 'accepted' immediately)")),
 		),
 		s.handleVMAction,
@@ -417,7 +417,7 @@ type vmAction struct {
 }
 
 // vmActions mirrors the CLI's vm action table (cmd/vm/action.go): ExpectStatus
-// and the destructive set (shutdown/force_shutdown/hibernate carry warnings
+// and the destructive set (shutdown/force_shutdown carry warnings
 // there; delete is special-cased) must stay in sync.
 var vmActions = map[string]vmAction{
 	verda.ActionStart: {
@@ -434,11 +434,6 @@ var vmActions = map[string]vmAction{
 		destructive:  true,
 		exec:         func(ctx context.Context, c *verda.Client, id string) error { return c.Instances.ForceShutdown(ctx, id) },
 	},
-	verda.ActionHibernate: {
-		expectStatus: verda.StatusOffline,
-		destructive:  true,
-		exec:         func(ctx context.Context, c *verda.Client, id string) error { return c.Instances.Hibernate(ctx, id) },
-	},
 	verda.ActionDelete: {
 		destructive: true,
 		exec: func(ctx context.Context, c *verda.Client, id string) error {

internal/verda-cli/cmd/vm/CLAUDE.md

@@ -2,7 +2,7 @@
 
 ## Quick Reference
 - Parent: `verda vm` (aliases: `instance`, `instances`)
-- Subcommands: `create`, `list` (alias `ls`), `describe`, `availability`, `action`, plus shortcuts (`start`, `shutdown`, `hibernate`, `delete`)
+- Subcommands: `create`, `list` (alias `ls`), `describe`, `availability`, `action`, plus shortcuts (`start`, `shutdown`, `delete`)
 - Files:
   - `vm.go` -- Parent command, registers subcommands and shortcuts
   - `create.go` -- Create command, flags, `createOptions` struct, request building, validation
@@ -59,7 +59,7 @@
 
 ### Action Availability
 - Start: only from `offline`
-- Shutdown / Force shutdown / Hibernate: only from `running`
+- Shutdown / Force shutdown: only from `running`
 - Delete: always available (no ValidFrom filter)
 - Delete sub-flow: fetches attached volumes, lets user multi-select which to delete, warns about continued billing for undeleted volumes
 

internal/verda-cli/cmd/vm/README.md

@@ -13,7 +13,6 @@ Parent command for creating and managing Verda Cloud VM instances (GPU and CPU).
 | `verda vm action` | Perform actions on a VM instance | `--id` |
 | `verda vm start` | Start an offline instance | `--all`, `--status`, `--hostname` |
 | `verda vm shutdown` | Shutdown a running instance | `--all`, `--status`, `--hostname` |
-| `verda vm hibernate` | Hibernate a running instance | `--all`, `--status`, `--hostname` |
 | `verda vm delete` | Delete an instance | `--all`, `--status`, `--hostname`, `--with-volumes` |
 
 Aliases: `verda instance`, `verda instances`
@@ -127,7 +126,6 @@ Interactive instance selector appears when `--id` is not provided. Available act
 | Start | offline |
 | Shutdown | running |
 | Force shutdown | running |
-| Hibernate | running |
 | Delete instance | always |
 
 Destructive actions (Shutdown, Force shutdown, Delete) show confirmation prompts with warnings. Delete has a sub-flow for selecting which attached volumes to also delete.
@@ -148,7 +146,7 @@ Destructive actions (Shutdown, Force shutdown, Delete) show confirmation prompts
 - **availability.go** -- `vm availability` with pricing table and spot price column
 - **action.go** -- `vm action` command, `instanceAction` data-driven action definitions, status-filtered action menu, confirmation flows, delete sub-flow with volume selection
 - **batch.go** -- Batch operations (`--all`, `--hostname` glob), per-instance error reporting, agent-mode structured JSON output
-- **shortcuts.go** -- Generated shortcut commands (`vm start`, `vm shutdown`, `vm hibernate`, `vm delete`) from `shortcutDef` structs
+- **shortcuts.go** -- Generated shortcut commands (`vm start`, `vm shutdown`, `vm delete`) from `shortcutDef` structs
 - **instances.go** -- Shared `fetchInstances` helper with spinner and status filtering, `filterByStatus`
 - **status_view.go** -- Instance card renderer, status-to-color mappings
 
@@ -157,7 +155,7 @@ Destructive actions (Shutdown, Force shutdown, Delete) show confirmation prompts
 - `client.Instances.Create` -- Create instance
 - `client.Instances.Get` -- List instances (with optional status filter)
 - `client.Instances.GetByID` -- Get instance details
-- `client.Instances.Start` / `Shutdown` / `ForceShutdown` / `Hibernate` -- Instance lifecycle
+- `client.Instances.Start` / `Shutdown` / `ForceShutdown` -- Instance lifecycle
 - `client.Instances.Delete` -- Delete instance with optional volume IDs
 - `client.Instances.Action` -- Bulk action on multiple instances
 - `client.InstanceAvailability.GetAllAvailabilities` -- Location/instance type availability

internal/verda-cli/cmd/vm/action.go

@@ -33,7 +33,6 @@ var actionNameMap = map[string]string{
 	verda.ActionShutdown:      "Shutdown",
 	verda.ActionForceShutdown: "Force shutdown",
 	"force-shutdown":          "Force shutdown", // CLI alias (hyphenated)
-	verda.ActionHibernate:     "Hibernate",
 	verda.ActionDelete:        "Delete instance",
 }
 
@@ -77,16 +76,6 @@ var allActions = []instanceAction{
 			return c.Instances.ForceShutdown(ctx, inst.ID)
 		},
 	},
-	{
-		Label:     "Hibernate",
-		ValidFrom: []string{verda.StatusRunning},
-		ConfirmMsg: "Hibernating the instance saves its state and stops billing.\n" +
-			"  You can resume it later.",
-		ExpectStatus: verda.StatusOffline,
-		Execute: func(ctx context.Context, c *verda.Client, inst *verda.Instance) error {
-			return c.Instances.Hibernate(ctx, inst.ID)
-		},
-	},
 	{
 		Label:   "Delete instance",
 		Execute: nil, // handled specially in runAction via runDeleteFlow
@@ -131,7 +120,7 @@ func NewCmdAction(f cmdutil.Factory, ioStreams cmdutil.IOStreams) *cobra.Command
 		Short: "Perform actions on a VM instance",
 		Long: cmdutil.LongDesc(`
 			Select a VM instance and perform an action: start, shutdown,
-			force shutdown, hibernate, or delete.
+			force shutdown, or delete.
 		`),
 		Example: cmdutil.Examples(`
 			# Interactive: select instance then action
@@ -154,7 +143,7 @@ func NewCmdAction(f cmdutil.Factory, ioStreams cmdutil.IOStreams) *cobra.Command
 	}
 
 	cmd.Flags().StringVar(&opts.InstanceID, "id", "", "Instance ID to act on")
-	cmd.Flags().StringVar(&opts.Action, "action", "", "Action to perform: start, shutdown, force_shutdown, hibernate, delete")
+	cmd.Flags().StringVar(&opts.Action, "action", "", "Action to perform: start, shutdown, force_shutdown, delete")
 	cmd.Flags().BoolVar(&opts.Yes, "yes", false, "Skip confirmation for destructive actions (required in agent mode)")
 	cmd.Flags().BoolVar(&opts.WithVolumes, "with-volumes", false, "Also delete all attached volumes (delete only); without it the OS volume survives detached and keeps billing")
 	// Hidden like on non-delete shortcuts: `vm delete --with-volumes` is the
@@ -575,7 +564,7 @@ func validFromForAction(actionName string) []string {
 func resolveAction(actionName string, validActions []instanceAction) (instanceAction, error) {
 	label, ok := actionNameMap[strings.ToLower(actionName)]
 	if !ok {
-		return instanceAction{}, fmt.Errorf("unknown --action %q: valid actions are start, shutdown, force_shutdown, hibernate, delete", actionName)
+		return instanceAction{}, fmt.Errorf("unknown --action %q: valid actions are start, shutdown, force_shutdown, delete", actionName)
 	}
 	for _, a := range validActions {
 		if a.Label == label {

internal/verda-cli/cmd/vm/action_test.go

@@ -58,9 +58,6 @@ func TestAvailableActions_Running(t *testing.T) {
 	if !labels["Force shutdown"] {
 		t.Error("expected Force shutdown action for running instances")
 	}
-	if !labels["Hibernate"] {
-		t.Error("expected Hibernate action for running instances")
-	}
 	if !labels["Delete instance"] {
 		t.Error("expected Delete action (always available)")
 	}

internal/verda-cli/cmd/vm/batch.go

@@ -277,8 +277,6 @@ func actionNameToAPI(action string) string {
 		return verda.ActionShutdown
 	case "force_shutdown", "force-shutdown":
 		return verda.ActionForceShutdown
-	case "hibernate":
-		return verda.ActionHibernate
 	case actionNameDelete:
 		return verda.ActionDelete
 	default:

internal/verda-cli/cmd/vm/batch_test.go

@@ -323,7 +323,7 @@ func TestShortcutCommandsHaveBatchFlags(t *testing.T) {
 	vmCmd := NewCmdVM(f, ioStreams)
 
 	// The shortcut commands to check. Some may be aliases.
-	wantCommands := []string{"shutdown", "start", "hibernate", "delete"}
+	wantCommands := []string{"shutdown", "start", "delete"}
 	for _, name := range wantCommands {
 		t.Run(name, func(t *testing.T) {
 			var found *cobra.Command
@@ -434,7 +434,7 @@ func TestHostnameFlag(t *testing.T) {
 
 	vmCmd := NewCmdVM(f, ioStreams)
 
-	for _, name := range []string{"shutdown", "start", "hibernate", "delete"} {
+	for _, name := range []string{"shutdown", "start", "delete"} {
 		t.Run(name, func(t *testing.T) {
 			var found *cobra.Command
 			for _, sub := range vmCmd.Commands() {
@@ -462,7 +462,6 @@ func TestActionNameToAPI(t *testing.T) {
 	}{
 		{"start", verda.ActionStart},
 		{"shutdown", verda.ActionShutdown},
-		{"hibernate", verda.ActionHibernate},
 		{"delete", verda.ActionDelete},
 		{"force_shutdown", verda.ActionForceShutdown},
 		{"force-shutdown", verda.ActionForceShutdown},

internal/verda-cli/cmd/vm/shortcuts.go

@@ -43,11 +43,6 @@ var shortcuts = []shortcutDef{
 		Short:  "Shutdown a VM instance",
 		Action: verda.ActionShutdown,
 	},
-	{
-		Use:    "hibernate <instance-id>",
-		Short:  "Hibernate a VM instance",
-		Action: verda.ActionHibernate,
-	},
 	{
 		Use:     "delete <instance-id>",
 		Aliases: []string{"rm"},