feat(cli): add interactive setup wizard and modernize init command

[cameroncooke]

Add xcodebuildmcp setup — an interactive terminal wizard that configures project defaults (project/workspace, scheme, simulator, workflows, debug mode, Sentry opt-out) and persists results to .xcodebuildmcp/config.yaml.

Previously users had to hand-edit YAML to configure their project. The setup wizard discovers available projects, schemes, and simulators, presents them in a polished terminal UI via @clack/prompts, and writes a complete config file with a summary of what changed.

Supporting changes:

• Shared Prompter abstraction (src/cli/interactive/prompts.ts) — wraps @clack/prompts with a testable interface that auto-selects defaults in non-interactive mode
• sentryDisabled promoted to first-class config key — previously env-var-only (XCODEBUILDMCP_SENTRY_DISABLED), now settable in config.yaml and resolved through the standard config layering. A new hydrateSentryDisabledEnvFromProjectConfig() bridges the config file to the env var before Sentry init in all three entrypoints (cli, daemon, mcp server)
• MCP tool refactors — extracted reusable discoverProjects(), listSchemes(), and listSimulators() functions from the MCP *Logic wrappers so both the MCP layer and CLI can call them directly. No behavior change to MCP tool responses
• init command modernized — switched from raw readline to @clack/prompts, added interactive skill type and client selection when no flags are provided, replaced Cursor/Codex targets with generic "Agents Skills" target, added ~ expansion for --dest
• persistProjectConfigPatch() — new utility for atomic config file updates with support for mutually exclusive key cleanup (projectPath vs workspacePath)

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/xcodebuildmcp@247

commit: 116725f

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

Bugbot Autofix prepared fixes for all 3 issues found in the latest run.

• ✅ Fixed: Interactive Claude selection silently skipped for MCP skill
  • Added check for Claude in user-selected targets array to treat interactive selection as explicit intent, same as --client flag.
• ✅ Fixed: Internal key filter misses camelCase logLevel variant
  • Added 'logLevel' to the internalKeys set to filter out the camelCase variant created by yargs camel-case-expansion.
• ✅ Fixed: Inconsistent TTY check in ensureAgentsGuidance after refactor
  • Changed guard from process.stdin.isTTY to isInteractiveTTY() to match the check used by promptConfirm function.

Or push these changes by commenting:

@cursor push 3034395277

Preview (3034395277)

diff --git a/src/cli/commands/init.ts b/src/cli/commands/init.ts
--- a/src/cli/commands/init.ts
+++ b/src/cli/commands/init.ts
@@ -286,7 +286,7 @@
   }
 
   if (!force) {
-    if (!process.stdin.isTTY) {
+    if (!isInteractiveTTY()) {
       throw new Error(
         `${AGENTS_FILE_NAME} exists and requires confirmation to update. Re-run with --force to apply the change in non-interactive mode.`,
       );
@@ -665,6 +665,11 @@
     return { allowedTargets: targets, skippedClients: [] };
   }
 
+  const hasClaudeTarget = targets.some((t) => t.id === 'claude');
+  if (clientFlag === 'claude' || hasClaudeTarget) {
+    return { allowedTargets: targets, skippedClients: [] };
+  }
+
   const allowedTargets: ClientInfo[] = [];
   const skippedClients: Array<{ client: string; reason: string }> = [];
 
@@ -679,9 +684,5 @@
     allowedTargets.push(target);
   }
 
-  if (clientFlag === 'claude') {
-    return { allowedTargets: targets, skippedClients: [] };
-  }
-
   return { allowedTargets, skippedClients };
 }

diff --git a/src/cli/register-tool-commands.ts b/src/cli/register-tool-commands.ts
--- a/src/cli/register-tool-commands.ts
+++ b/src/cli/register-tool-commands.ts
@@ -162,7 +162,16 @@
 
       // Convert CLI argv to tool params (kebab-case -> camelCase)
       // Filter out internal CLI options before converting
-      const internalKeys = new Set(['json', 'output', 'style', 'socket', 'log-level', '_', '$0']);
+      const internalKeys = new Set([
+        'json',
+        'output',
+        'style',
+        'socket',
+        'log-level',
+        'logLevel',
+        '_',
+        '$0',
+      ]);
       const flagArgs: Record<string, unknown> = {};
       for (const [key, value] of Object.entries(argv as Record<string, unknown>)) {
         if (!internalKeys.has(key)) {

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

[sentry]

Bug: The enforceInstallPolicy function incorrectly bypasses MCP skill filtering for the 'claude' client when selectionMode is 'interactive', installing an unnecessary skill.
_{Severity: MEDIUM}

Suggested Fix

Remove the || selectionMode === 'interactive' condition from the if statement within the enforceInstallPolicy function. This will ensure the filtering logic that excludes Claude for MCP skills is applied consistently across all selection modes.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: src/cli/commands/init.ts#L683-L689

Potential issue: The `enforceInstallPolicy` function is designed to prevent the
installation of 'mcp' skills for the 'claude' client, as they are unnecessary. However,
a conditional check `if (clientFlag === 'claude' || selectionMode === 'interactive')`
bypasses this filtering logic when a user selects clients in interactive mode. This
results in the function returning the original, unfiltered `targets` array, leading to
the installation of an unnecessary skill for Claude. This scenario is not covered by
existing tests.

_{Did we get this right? 👍 / 👎 to inform future reviews.}