pin all CLI installations to explicit versions and use pro models by default (#283)

* pin all CLI installations to explicit versions and use pro models by default

- codex: pin @openai/codex to 0.101.0 (was "latest")
- opencode: pin opencode-ai to 1.1.56 (was "latest")
- gemini: pin gemini-cli to v0.28.2 via new tag param on installFromGithub
- cursor: pin to 2026.01.28-fd13201 via direct tarball download (replaces curl install script)
- add installFromDirectTarball to install.ts for versioned tarball URLs
- gemini auto effort now uses pro-preview instead of flash-preview
- test runner model overrides updated to use pro-preview consistently

Co-authored-by: Cursor <cursoragent@cursor.com>

* increase activity timeout

* fix delegation timeout

---------

Co-authored-by: Cursor <cursoragent@cursor.com>

agents/gemini.ts

@@ -24,7 +24,7 @@ const geminiEffortConfig: Record<Effort, { model: string; thinkingLevel: string
   // pass the model directly it works if we ever did need to do something like this,
   // we could write to .gemini/settings.json
   mini: { model: "gemini-3-flash-preview", thinkingLevel: "LOW" },
-  auto: { model: "gemini-3-flash-preview", thinkingLevel: "HIGH" },
+  auto: { model: "gemini-3-pro-preview", thinkingLevel: "HIGH" },
   max: { model: "gemini-3-pro-preview", thinkingLevel: "HIGH" },
 } as const;
 
@@ -85,6 +85,9 @@ type GeminiEvent =
   | GeminiToolResultEvent
   | GeminiResultEvent;
 
+// pinned CLI version — gemini-cli is installed from GitHub releases, not npm
+const GEMINI_CLI_VERSION = "v0.28.2";
+
 // transient API error patterns that warrant a retry.
 // these are server-side issues, not client errors.
 const TRANSIENT_ERROR_PATTERNS = [
@@ -191,6 +194,7 @@ async function installGemini(githubInstallationToken?: string): Promise<string>
   return await installFromGithub({
     owner: "google-gemini",
     repo: "gemini-cli",
+    tag: GEMINI_CLI_VERSION,
     assetName: "gemini.js",
     ...(githubInstallationToken && { githubInstallationToken }),
   });
@@ -231,6 +235,7 @@ export const gemini = agent({
           cmd: "node",
           args: [cliPath, ...args],
           env: process.env,
+          activityTimeout: 0, // disabled: process-level timeout in main.ts handles this (subprocess timeout would kill orchestrator during delegation)
           onStdout: async (chunk) => {
             const text = chunk.toString();
             finalOutput += text;