@proagentstore/cli 0.3.7 → 0.3.8

package/dist/browser-runner/browser-actions.js

@@ -63,6 +63,11 @@ async function typeRobustly(loc, text) {
  */
 export async function performBrowserAction(page, action, clickRobustly, resumeFile) {
     const locate = () => {
+        // Prefer the stable snapshot ref (aria-ref) — points at the exact element, so two
+        // fields sharing a label (e.g. a phone "Country" code and an address "Country")
+        // are never confused. Fall back to role+name for older snapshots.
+        if (action.ref)
+            return page.locator(`aria-ref=${action.ref}`);
         const role = action.role;
         let loc = role
             ? page.getByRole(role, action.name ? { name: action.name } : undefined)
@@ -267,12 +272,17 @@ export async function performBrowserAction(page, action, clickRobustly, resumeFi
 export async function inspectField(page, action) {
     if (action.action !== "type" && action.action !== "select" && action.action !== "check")
         return "";
-    if (!action.name)
+    if (!action.name && !action.ref)
         return "";
     try {
         const role = action.role;
-        let loc = role ? page.getByRole(role, { name: action.name }) : page.getByText(action.name, { exact: false });
-        loc = typeof action.nth === "number" ? loc.nth(action.nth) : loc.first();
+        let loc = action.ref
+            ? page.locator(`aria-ref=${action.ref}`)
+            : role
+                ? page.getByRole(role, action.name ? { name: action.name } : undefined)
+                : page.getByText(action.name ?? "", { exact: false });
+        if (!action.ref)
+            loc = typeof action.nth === "number" ? loc.nth(action.nth) : loc.first();
         const el = await loc.elementHandle({ timeout: 1_500 }).catch(() => null);
         if (!el)
             return "";

package/dist/browser-runner/runner.js

@@ -620,7 +620,15 @@ export class LocalRunner {
         }
         const page = await this.getActivePage();
         await page.waitForLoadState("domcontentloaded", { timeout: 5_000 }).catch(() => undefined);
-        const full = await page.locator("body").ariaSnapshot().catch(() => "");
+        // Match what a direct Claude + Playwright-MCP agent sees: `ariaSnapshot({mode:"ai"})`
+        // carries element STATE (disabled/checked/expanded/selected/value) and a stable
+        // [ref=eNN] on each interactive element, so the brain targets them unambiguously via
+        // the aria-ref selector (instead of role+name, which collides when two fields share a
+        // label). This is exactly the call Playwright's own MCP uses. Falls back to the plain
+        // aria snapshot if the option is ever unavailable.
+        const aria = page;
+        const full = (await aria.ariaSnapshot({ mode: "ai" }).catch(() => "")) ||
+            (await page.locator("body").ariaSnapshot().catch(() => ""));
         const MAX = 16_000;
         const snapshot = full.length > MAX ? `${full.slice(0, MAX)}\n… [snapshot truncated]` : full;
         return {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@proagentstore/cli",
-  "version": "0.3.7",
+  "version": "0.3.8",
   "description": "CLI for creating, publishing, and running ProAgentStore agents",
   "license": "MIT",
   "type": "module",