Skip to content

Conversation

@patricebechard
Copy link
Collaborator

Summary

  • Introduces a comprehensive set of new "Dynamic Guidance" tasks for evaluating web agents on ServiceNow workflows
  • Adds 12+ new task categories covering various ServiceNow operations
  • Includes 30+ JSON configuration files with diverse task samples and examples

New Task Categories

Case Management

  • GetCaseStatusTask - Query case status
  • CloseCaseTask - Close cases with resolution codes
  • FindAssetUnderAccountCreateCaseTask - Complex multi-step case creation

User & Role Management

  • AssignRoleToUserAdminTask - Assign admin roles
  • AssignRolesToUserExplicitTask - Explicit role assignments
  • AssignRolesToUserImplicitTask - Implicit role assignments
  • CreateUserGroupTask / DeactivateUserGroupTask - User group lifecycle

Incident Management

  • AddAdditionalAssigneeToIncidentTask
  • ResolveIncidentTask
  • UpdateIncidentTask
  • CreateWorkspaceIncidentTask - Workspace-specific incident creation

Service Catalog (Dynamic Guidance)

  • Order iPhone, mobile phones, software, hardware
  • Order with business justification
  • Reset password requests
  • Packaging and shipping orders

Other Tasks

  • Change request management
  • License queries
  • RITM (Request Item) status and quantity updates
  • Interaction creation
  • Customer account lookups
  • Navigation with custom goals
  • Workspace form tasks (incident/problem creation)

Technical Changes

  • Added ALL_WORKARENA_DYNAMIC_GUIDANCE_TASKS task registry in init.py
  • New "dg" filter level in get_all_tasks_agents() for dynamic guidance task selection
  • Extended TASK_CATEGORY_MAP with new service catalog task mappings
  • New configuration paths in config.py for all task types

Test Plan

  • Verify all new tasks are properly registered with BrowserGym
  • Run sample tasks from each category to validate setup/teardown
  • Confirm validators correctly evaluate task completion
  • Test task sampling with filter="dg" option

@patricebechard patricebechard self-assigned this Jan 14, 2026
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.

4 participants