Keep Ask AI conversations intact through agent action setup

[Paul Lizer (paullizer)]

Summary

• Users building an agent with Ask AI previously had to find drafted actions in the Actions section, then lost their conversation after finishing setup. Inline cards now show missing details and open the existing action workbench directly.
• The workbench uses Back to agent and Save action and return to agent. Returning restores the conversation, unsent message, panel state and reading position, with a confirmation that the action is selected but the agent remains unsaved.
• An explicit, bounded, owner/workspace-scoped in-memory handoff preserves the same editor visit instead of reusing history for all reopened drafts. Refreshes and unrelated visits remain fresh; existing permissions, discard warnings and conflict-safe Undo are preserved.
• Adds SQL desktop/mobile and scoped-editor regression coverage, updates the agent/action guides and fix documentation, and bumps the application to 0.261.333. No new configuration form, credential collection in Ask AI, or persistent conversation storage.

Linked issue

N/A. The user requested a local plan without creating an issue.

Release Notes & Latest Features

• New Feature
• Bug Fix
• UI Enhancement
• Breaking Change
• Internal only

Is this visible to end users?

• Yes
• No

Is this admin-facing (Admin Settings, governance, deployment, config)?

• Yes
• No

The shared editor flow also supports administrator-owned global agents/actions. Governance and deployment behavior are unchanged.

Should this become a Latest Feature card?

• Yes
• No
• Already added

Screenshot needed for the card?

• Yes
• No
• Attached

Version bump

• application/single_app/config.py VERSION third segment bumped, or not needed because this is docs-only
• deployers/version.txt bumped, or not needed because deployers/ was not changed

Application version: 0.261.332 -> 0.261.333. No deployer changes.

Testing / validation

• npm --prefix .\application\v2_ui run build - passed TypeScript compilation and production build.
• node .\functional_tests\test_v2_editor_assist_handoff.mjs - passed lifecycle, bounded retention, abandonment and owner/workspace isolation checks.
• node .\functional_tests\test_v2_agent_editor_assist_logic.mjs - passed.
• node .\functional_tests\test_v2_action_editor_assist_logic.mjs - 7/7 passed.
• python .\functional_tests\test_v2_assist_thread.py - 14/14 passed, including 140 shared TypeScript logic checks.
• python .\scripts\build_docs_inventory.py - regenerated successfully; inventory contents unchanged.
• python .\functional_tests\test_docs_app_surface_coverage.py - 7/7 passed.
• python .\functional_tests\test_docs_site_quality.py - 6/6 passed.

The following production-bundle browser run passed 41 tests:

$env:PLAYWRIGHT_SERVICE_URL = ''; $env:PYTHONPATH = '.\ui_tests\fixtures;.\functional_tests;.\application\single_app'; python -m pytest .\ui_tests\test_v2_agent_action_ask_ai.py .\ui_tests\test_v2_admin_global_agents_actions.py .\ui_tests\test_v2_workspace_authoring.py::test_new_action_roundtrip_retains_agent_draft_without_premature_save .\ui_tests\test_v2_workspace_authoring.py::test_cancelled_new_action_returns_unchanged_unsaved_agent .\ui_tests\test_v2_workspace_authoring.py::test_unsafe_action_return_path_stays_inside_workspace .\ui_tests\test_v2_workspace_authoring.py::test_dirty_cancel_workspace_navigation_back_forward_and_reload .\ui_tests\test_v2_group_agents.py::test_new_group_action_from_the_agent_editor_returns_into_the_retained_draft -q --tb=short

Coverage includes inline SQL setup on desktop/mobile, restored follow-up context and composer, keyboard/focus behavior, Back/Cancel/browser Back, save failure and retry, multiple/renamed actions, safe Undo after saving, and personal/group/global return paths. Desktop/mobile screenshots were inspected locally; no live Azure service, SQL database or language model was contacted.

The added-line XSS sink scan passed. A full-file scan additionally flags the pre-existing Link to={adapter.basePath} in the agent editor's error fallback; that unchanged trusted-adapter link is outside this fix.

Documentation

• Release notes updated, or not needed
• Feature documentation updated, or not needed
• Fix documentation updated, or not needed

Updated docs/guides/create-an-agent.md and docs/guides/create-an-action.md. Added docs/explanation/fixes/ASK_AI_ACTION_HANDOFF_CONTINUITY_FIX.md with root cause, boundaries and validation results. Release notes were offered but remain unchanged pending user approval; no Latest Feature card was added.

Security checklist

• New Flask routes include @swagger_route(security=get_auth_security())
• Settings sent to non-admin frontends use sanitize_settings_for_user()
• Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS
• No secrets, keys, connection strings, or local-only artifacts are included

No route or settings-response changes. Cards use React text nodes and real pending-action identifiers; navigation reuses validated return paths. Credentials remain in the existing action editor, not the handoff transcript metadata. Saving an action never silently saves its agent, and Undo never deletes the separately saved action.