Skip to content

Consolidate browser automation on WebFlows; replace direct actions wi…#2093

Merged
hillary-mutisya merged 1 commit intomicrosoft:mainfrom
hillary-mutisya:authoring11
Mar 28, 2026
Merged

Consolidate browser automation on WebFlows; replace direct actions wi…#2093
hillary-mutisya merged 1 commit intomicrosoft:mainfrom
hillary-mutisya:authoring11

Conversation

@hillary-mutisya
Copy link
Copy Markdown
Collaborator

…th draft scripts

  • Remove 9 direct browser action types (clickOnElement, enterTextInElement, setDropdownValue, awaitPageLoad, getHTML, getElementByDescription, isPageStateMatched, queryPageContent, downloadImage) and their handlers from the browser agent. These were previously used by reasoning agents but are now superseded by WebFlow-based execution.

  • Add checkPageState() and queryContent() to the WebFlow browser API, preserving the LLM-based semantic query capabilities that were unique to the removed actions.

  • Replace the webtask recipe generator (TaskFlow recipes) with a WebFlow generator that produces WebFlowDefinitions from execution plans. Add webflow__list, webflow__execute, and webflow__run_draft MCP tools so the webtask agent can discover, invoke, and author WebFlows. Route webflow schema calls through the execute_action handler to the browser agent. Remove broken MCP tools that referenced deleted actions.

  • Update the webtask subagent prompt to use draft scripts for all page interactions instead of direct DOM tools or execute_command. Add integration test tasks for the WebFlow-based execution model.

…th draft scripts

- Remove 9 direct browser action types (clickOnElement, enterTextInElement, setDropdownValue, awaitPageLoad, getHTML, getElementByDescription, isPageStateMatched, queryPageContent, downloadImage) and their handlers from the browser agent. These were previously used by reasoning agents but are now superseded by WebFlow-based execution.

- Add checkPageState() and queryContent() to the WebFlow browser API, preserving the LLM-based semantic query capabilities that were unique to the removed actions.

- Replace the webtask recipe generator (TaskFlow recipes) with a WebFlow generator that produces WebFlowDefinitions from execution plans. Add webflow__list, webflow__execute, and webflow__run_draft MCP tools so the webtask agent can discover, invoke, and author WebFlows. Route webflow schema calls through the execute_action handler to the browser agent. Remove broken MCP tools that referenced deleted actions.

- Update the webtask subagent prompt to use draft scripts for all page interactions instead of direct DOM tools or execute_command. Add integration test tasks for the WebFlow-based execution model.
@hillary-mutisya hillary-mutisya added this pull request to the merge queue Mar 28, 2026
Merged via the queue into microsoft:main with commit 378a723 Mar 28, 2026
13 of 15 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant