add phone number format checking (#2125)

skyvern/forge/prompts/skyvern/check-phone-number-format.j2

@@ -0,0 +1,41 @@
+You need to help checking if the current phone number format is matching the required format according to the user goal, user details and HTML elements.
+
+There are several checkpoints to verify the format of the phone number:
+- Whether the phone number should add hyphen.
+- Whether the phone number should add country code.
+- Whether the phone number should add space between the country code and the area code.
+
+MAKE SURE YOU OUTPUT VALID JSON. No text before or after JSON, no trailing commas, no comments (//), no unnecessary quotes, etc.
+
+Reply in JSON format with the following keys:
+{
+"phone_number_format": str, // Think step by step. The format of the phone number required on the page according to HTML elements.
+"thought": str, // Think step by step. Describe your thought about how you come up with the phone_number_format. Use information you see on the site to explain.
+"is_current_format_correct": bool, // True if the current phone number format is matching the required format.
+"recommended_phone_number": str, // If is_current_format_correct is True, return null. Otherwise, return the recommended phone number with the correct format.
+}
+
+Current phone number:
+```
+{{ current_phone_number }}
+```
+
+User goal:
+```
+{{ navigation_goal }}
+```
+
+User details:
+```
+{{ navigation_payload_str }}
+```
+
+HTML elements:
+```
+{{ elements }}
+```
+
+Current datetime, ISO format:
+```
+{{ local_datetime }}
+```

skyvern/utils/prompt_engine.py

@@ -1,6 +1,7 @@
 from typing import Any
 
 import structlog
+from pydantic import BaseModel
 
 from skyvern.constants import DEFAULT_MAX_TOKENS
 from skyvern.forge.sdk.prompting import PromptEngine
@@ -10,6 +11,13 @@ from skyvern.webeye.scraper.scraper import ScrapedPage
 LOG = structlog.get_logger()
 
 
+class CheckPhoneNumberFormatResponse(BaseModel):
+    phone_number_format: str
+    thought: str
+    is_current_format_correct: bool
+    recommended_phone_number: str | None
+
+
 def load_prompt_with_elements(
     scraped_page: ScrapedPage,
     prompt_engine: PromptEngine,

skyvern/webeye/actions/handler.py

@@ -68,7 +68,7 @@ from skyvern.forge.sdk.db.enums import OrganizationAuthTokenType
 from skyvern.forge.sdk.models import Step
 from skyvern.forge.sdk.schemas.tasks import Task
 from skyvern.forge.sdk.services.bitwarden import BitwardenConstants
-from skyvern.utils.prompt_engine import load_prompt_with_elements
+from skyvern.utils.prompt_engine import CheckPhoneNumberFormatResponse, load_prompt_with_elements
 from skyvern.webeye.actions import actions
 from skyvern.webeye.actions.actions import (
     Action,
@@ -211,6 +211,52 @@ def clean_and_remove_element_tree_factory(
     return helper_func
 
 
+async def check_phone_number_format(
+    phone_number: str,
+    action: actions.InputTextAction,
+    skyvern_element: SkyvernElement,
+    scraped_page: ScrapedPage,
+    task: Task,
+    step: Step,
+) -> str:
+    # check the phone number format
+    LOG.info(
+        "Input is a tel input, trigger phone number format checking",
+        action=action,
+        element_id=skyvern_element.get_id(),
+    )
+
+    new_scraped_page = await scraped_page.generate_scraped_page_without_screenshots()
+    html = new_scraped_page.build_element_tree(html_need_skyvern_attrs=False)
+    prompt = prompt_engine.load_prompt(
+        template="check-phone-number-format",
+        current_phone_number=phone_number,
+        navigation_goal=task.navigation_goal,
+        navigation_payload_str=json.dumps(task.navigation_payload),
+        elements=html,
+        local_datetime=datetime.now(skyvern_context.ensure_context().tz_info).isoformat(),
+    )
+
+    json_response = await app.SECONDARY_LLM_API_HANDLER(
+        prompt=prompt, step=step, prompt_name="check-phone-number-format"
+    )
+
+    check_phone_number_format_response = CheckPhoneNumberFormatResponse.model_validate(json_response)
+    if (
+        check_phone_number_format_response.is_current_format_correct
+        or not check_phone_number_format_response.recommended_phone_number
+    ):
+        return phone_number
+
+    LOG.info(
+        "The current phone number format is incorrect, using the recommended phone number",
+        action=action,
+        element_id=skyvern_element.get_id(),
+        recommended_phone_number=check_phone_number_format_response.recommended_phone_number,
+    )
+    return check_phone_number_format_response.recommended_phone_number
+
+
 class AutoCompletionResult(BaseModel):
     auto_completion_attempt: bool = False
     incremental_elements: list[dict] = []
@@ -557,7 +603,7 @@ async def handle_input_text_action(
 
     # before filling text, we need to validate if the element can be filled if it's not one of COMMON_INPUT_TAGS
     tag_name = scraped_page.id_to_element_dict[action.element_id]["tagName"].lower()
-    text = await get_actual_value_of_parameter_if_secret(task, action.text)
+    text: str | None = await get_actual_value_of_parameter_if_secret(task, action.text)
     if text is None:
         return [ActionFailure(FailedToFetchSecret())]
 
@@ -706,6 +752,25 @@ async def handle_input_text_action(
 
     # force to move focus back to the element
     await skyvern_element.get_locator().focus(timeout=timeout)
+
+    # check the phone number format
+    if await skyvern_element.get_attr("type") == "tel":
+        try:
+            text = await check_phone_number_format(
+                phone_number=text,
+                action=action,
+                skyvern_element=skyvern_element,
+                scraped_page=scraped_page,
+                task=task,
+                step=step,
+            )
+        except Exception:
+            LOG.warning(
+                "Failed to check the phone number format, using the original text",
+                action=action,
+                exc_info=True,
+            )
+
     # `Locator.clear()` on a spin button could cause the cursor moving away, and never be back
     if not await skyvern_element.is_spinbtn_input():
         try: