go.mod (2)

88-88: Use the canonical YAML module path.

go.yaml.in/yaml/v3 exists, but the canonical and widely used path is gopkg.in/yaml.v3. Prefer the canonical path to avoid tooling surprises.

Apply this diff:

-	go.yaml.in/yaml/v3 v3.0.4 // indirect
+	gopkg.in/yaml.v3 v3.0.1 // indirect

Reference: official import path is gopkg.in/yaml.v3. (github.com)

---

15-15: Optional: avoid long‑lived pseudo‑versions for tools/UI libs.

If stability matters, consider pinning github.com/charmbracelet/huh/spinner to a tagged release once available to reduce churn from pseudo‑versions. This is optional.

internal/common.go (2)

53-55: Remove duplicate sorting.

GetRegions already sorts regionNames. The extra sort in SelectRegion is redundant.

Apply this diff:

-	sort.Slice(regionNames, func(i, j int) bool {
-		return regionNames[i] < regionNames[j]
-	})

Also applies to: 36-39

---

69-77: Handle cancellation robustly; avoid string‑matching “^C”.

Compare with context.Canceled (or ctx.Err()) instead of matching error text.

Apply this diff:

-		select {
-		case <-ctx.Done():
-			return "", ctx.Err()
-		default:
-			if err.Error() == "^C" {
-				return "", context.Canceled
-			}
-			return "", fmt.Errorf("failed to select region: %w", err)
-		}
+		if errors.Is(err, context.Canceled) || ctx.Err() != nil {
+			return "", context.Canceled
+		}
+		return "", fmt.Errorf("failed to select region: %w", err)

tailout/connect.go (1)

49-51: Optional: auto‑select when only one node is available.

If exactly one active node exists and interactive mode is enabled, select it without prompting for a smoother UX.

tailout/create.go (2)

291-293: Avoid returning (nil, nil) — an anti-pattern in Go.

When the user declines instance creation, the function returns (nil, nil). This requires the caller to check both the return value and error, which is error-prone. Instead, return a sentinel error or a custom error type to clearly indicate user cancellation.

Define a sentinel error at the package level:

var ErrUserCancelled = errors.New("user cancelled operation")

Then apply this diff:

 	if !result {
-		return nil, nil
+		return nil, ErrUserCancelled
 	}

Update the caller (lines 102-109) to check for this error:

 	runInput, errPrep := prepareInstance(ctx, cfg, aws.Bool(dryRun))
 	if errPrep != nil {
+		if errors.Is(errPrep, ErrUserCancelled) {
+			fmt.Println("Instance creation cancelled.")
+			return nil
+		}
 		return fmt.Errorf("failed to prepare instance: %w", errPrep)
 	}
-	if runInput == nil {
-		fmt.Println("Instance creation aborted.")
-		return nil
-	}

---

363-370: Use a more descriptive variable name.

The variable instance1 is confusing. Consider renaming it to describedInstance or runningInstance to clarify that it's the instance retrieved from the DescribeInstances call.

Apply this diff:

-	instance1 := reservation.Instances[0]
-	if instance1.PublicIpAddress == nil {
+	describedInstance := reservation.Instances[0]
+	if describedInstance.PublicIpAddress == nil {
 		return instance, errors.New("no public IP address found")
 	}
 
 	instance.Name = nodeName
 	instance.InstanceID = *createdInstance.InstanceId
-	instance.IP = *instance1.PublicIpAddress
+	instance.IP = *describedInstance.PublicIpAddress

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

tailout/connect.go (2)

54-60: Avoid panic on empty addresses; add defensive check.

device.Addresses[0] will panic if the slice is empty. Add a guard similar to the pattern used in tailout/stop.go lines 51-57.

Apply this diff:

 		// Create options for huh select
 		options := make([]huh.Option[int], len(tailoutDevices))
 		for i, device := range tailoutDevices {
-			// Display hostname and IP address
-			label := fmt.Sprintf("%s (%s)", device.Hostname, device.Addresses[0])
+			addr := "no IP"
+			if len(device.Addresses) > 0 {
+				addr = device.Addresses[0]
+			}
+			label := fmt.Sprintf("%s (%s)", device.Hostname, addr)
 			options[i] = huh.NewOption(label, i)
 		}

---

88-88: Avoid panic on empty addresses in success message.

Same issue as lines 54-60: deviceToConnectTo.Addresses[0] can panic if Addresses is empty.

Apply this diff:

+	addr := "no IP"
+	if len(deviceToConnectTo.Addresses) > 0 {
+		addr = deviceToConnectTo.Addresses[0]
+	}
-	fmt.Printf("Connected to node %s (%s) via Tailscale.\n", deviceToConnectTo.Hostname, deviceToConnectTo.Addresses[0])
+	fmt.Printf("Connected to node %s (%s) via Tailscale.\n", deviceToConnectTo.Hostname, addr)

tailout/create.go (2)

243-243: Use the calculated durationMinutes variable instead of hardcoded value.

The shutdown time is hardcoded as 10 minutes, but the durationMinutes variable (calculated from the user-provided shutdown config at lines 84-87) is never used.

Apply this diff:

-sudo echo "sudo shutdown" | at now + ` + strconv.Itoa(10) + ` minutes`
+sudo echo "sudo shutdown" | at now + ` + strconv.Itoa(durationMinutes) + ` minutes`

Note: You'll need to pass durationMinutes as a parameter to prepareInstance to access it here.

---

290-290: Display the actual shutdown duration instead of "TODO".

The user-facing output shows "TODO" for the auto shutdown time. Pass durationMinutes as a parameter to prepareInstance to display the actual value.

Apply this diff:

 - Auto shutdown after: %s
-	`, *identity.Account, imageID, imageName, imageOwner, imageArchitecture, types.InstanceTypeT3aMicro, cfg.Region, "TODO")
+	`, *identity.Account, imageID, imageName, imageOwner, imageArchitecture, types.InstanceTypeT3aMicro, cfg.Region, strconv.Itoa(durationMinutes) + " minutes")

And update the function signature:

-func prepareInstance(ctx context.Context, cfg aws.Config, dryRun *bool) (instance *ec2.RunInstancesInput, err error) {
+func prepareInstance(ctx context.Context, cfg aws.Config, dryRun *bool, durationMinutes int) (instance *ec2.RunInstancesInput, err error) {

Then update the call at line 104:

-	runInput, errPrep := prepareInstance(ctx, cfg, aws.Bool(dryRun))
+	runInput, errPrep := prepareInstance(ctx, cfg, aws.Bool(dryRun), durationMinutes)

internal/common.go (2)

53-55: Remove redundant sort.

GetRegions already sorts the region names at lines 36-38, so this sort is unnecessary.

Apply this diff:

-	sort.Slice(regionNames, func(i, j int) bool {
-		return regionNames[i] < regionNames[j]
-	})
-
 	var selectedRegion string

---

90-93: Add context cancellation check for consistency.

For consistency with SelectRegion (lines 69-71), check for context cancellation explicitly.

Apply this diff:

 	err := form.RunWithContext(ctx)
 	if err != nil {
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return false, fmt.Errorf("prompt canceled: %w", ctxErr)
+		}
 		return false, fmt.Errorf("failed to prompt for yes/no: %w", err)
 	}

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

tailout/create.go (3)

138-139: Clarify or remove the dry run comment.

The comment "// If dry run, exit here?" is unclear because dry run is already handled within createInstance at lines 310-313. If additional dry run logic is needed here, implement it; otherwise, remove the comment.

Apply this diff if no additional logic is needed:

-	// If dry run, exit here?
-

---

282-290: Consider clarifying the shutdown duration unit in the display.

The display shows shutdownDuration but doesn't explicitly state the unit. Consider adding " minutes" to line 288 for clarity.

Apply this diff:

-- Auto shutdown after: %s
+- Auto shutdown after: %s minutes

---

344-346: Consider increasing the instance status wait timeout.

The 5-minute timeout for InstanceStatusOkWaiter might be insufficient for spot instances, which can take longer to reach running status, especially during high demand periods. Consider increasing to 10 minutes.

Apply this diff:

-	}, time.Minute*5)
+	}, time.Minute*10)

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

tailout/connect.go (1)

93-98: Previous issue resolved: defensive address handling in success message.

Lines 94-97 apply the same safe pattern as the interactive flow, checking array length before accessing deviceToConnectTo.Addresses[0]. This completes the mitigation of the panic risk flagged in earlier reviews.

Optional: The address fallback logic appears twice (lines 58-61 and here). Extracting it into a helper function like getDeviceAddress(device tsapi.Device) string would reduce duplication, but given the minimal size (3 lines), this is not essential.

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

internal/common.go (1)

170-176: Make post-update verification robust when ExitNodeStatus is nil.

If the node isn’t in the current netmap, ExitNodeStatus can be nil even though the policy was applied; confirm via prefs before failing. This was raised earlier and remains unaddressed.

- status, err = localClient.Status(ctx)
+ status, err = localClient.Status(ctx)
  if err != nil {
    return fmt.Errorf("failed to get tailscale status: %w", err)
  }
- if id != "" && (status.ExitNodeStatus == nil || status.ExitNodeStatus.ID != tailcfg.StableNodeID(id)) {
-   return errors.New("failed to set the exit node")
- }
+ // Re-fetch prefs to validate policy application.
+ postPrefs, err := localClient.GetPrefs(ctx)
+ if err != nil {
+   return fmt.Errorf("failed to get prefs after update: %w", err)
+ }
+ if id != "" {
+   if status.ExitNodeStatus != nil {
+     if status.ExitNodeStatus.ID != tailcfg.StableNodeID(id) {
+       return fmt.Errorf("exit node mismatch: want %s, got %s", tailcfg.StableNodeID(id), status.ExitNodeStatus.ID)
+     }
+   } else if postPrefs.ExitNodeID != tailcfg.StableNodeID(id) {
+     return errors.New("exit node ID not applied")
+   }
+ } else {
+   if postPrefs.ExitNodeID != "" || status.ExitNodeStatus != nil {
+     return errors.New("failed to clear the exit node")
+   }
+ }

internal/common.go (7)

36-38: Use sort.Strings for simpler intent.

Directly sort the slice of strings.

- sort.Slice(regionNames, func(i, j int) bool {
-   return regionNames[i] < regionNames[j]
- })
+ sort.Strings(regionNames)

---

43-44: Fix exported doc comment style.

Make the comment start with the function name to satisfy linters.

-// Function that uses huh to return an AWS region fetched from the aws sdk.
+// SelectRegion uses huh to return an AWS region fetched via the AWS SDK.

---

74-76: LGTM: context-aware yes/no prompt.

Minor: mirror doc comment style like above for linters.

-// Function that uses huh to return a boolean value.
+// PromptYesNo uses huh to return a boolean value.

Also applies to: 86-95

---

125-127: Prefer ipn.Running constant; include actual state in error.

Avoid string literal comparison.

- if status.BackendState != "Running" {
-   return errors.New("tailscale is not running")
- }
+ if status.BackendState != ipn.Running {
+   return fmt.Errorf("tailscale backend not running (state=%s)", status.BackendState)
+ }

---

150-166: Minimize MaskedPrefs to avoid stomping unrelated Prefs.

Update only the field you intend to change.

- prefs, err := localClient.GetPrefs(ctx)
- if err != nil {
-   return fmt.Errorf("failed to get prefs: %w", err)
- }
-
- if id != "" {
-   fmt.Printf("Setting exit node to %s...\n", id)
-   prefs.ExitNodeID = tailcfg.StableNodeID(id)
- } else {
-   fmt.Println("Clearing exit node...")
-   prefs.ClearExitNode()
- }
- _, err = localClient.EditPrefs(ctx, &ipn.MaskedPrefs{
-   Prefs:         *prefs,
-   ExitNodeIDSet: true,
- })
+ if id != "" {
+   fmt.Printf("Setting exit node to %s...\n", id)
+   _, err = localClient.EditPrefs(ctx, &ipn.MaskedPrefs{
+     Prefs:         ipn.Prefs{ExitNodeID: tailcfg.StableNodeID(id)},
+     ExitNodeIDSet: true,
+   })
+ } else {
+   fmt.Println("Clearing exit node...")
+   _, err = localClient.EditPrefs(ctx, &ipn.MaskedPrefs{
+     // zero value clears ExitNodeID when ExitNodeIDSet is true
+     Prefs:         ipn.Prefs{},
+     ExitNodeIDSet: true,
+   })
+ }
  if err != nil {
    return fmt.Errorf("failed to set/unset exit node: %w", err)
  }

---

145-147: Avoid printing from internal packages.

Return status strings or accept an io.Writer/logger so CLI decides how to display.

- fmt.Printf("Currently connected to exit node: %s\n", currentExitNodeName)
+ // return this info to caller for display

Option: change signature to accept an io.Writer or a logger interface.

Also applies to: 155-161

---

178-180: Reachability check is good; consider downgraded severity.

If policy was applied but node is offline, consider returning a typed error so callers can choose between warn/fail.

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro