[jhomra21]

Problem

Factory Droid sets each subagent's effort from the Task tool's complexity and reports it, but the child's model selection has no effort option. The hover card reads effort only from the model selection, so it shows the model without the effort.

Change

subagentRuntime.ts keeps the effort the provider reported for a child. SubagentTooltipContent falls back to it when the model selection has no effort, and its four callers pass it: ThreadLineageWorkflowRow, ThreadRelationshipsControl, V2LifecycleRow and WorkflowOpeningBlock. An effort in the model selection still wins. The optional contract field that carries the reported effort comes from the native provider PR.

Scope and approval

This depends on the native Factory Droid provider PR and falls under the same Ideas discussion. (Ideas discussion: pingdotgg#27369) It only adds the fallback. The diff is 7 files against the provider branch, and it doesn't include the sidebar PR.

Verification

• vp env exec --node 26.12.1 -- vp test run packages/client-runtime/src/state/subagentRuntime.test.ts: 12 passed, including a case where the runtime keeps the reported effort.
• vp env exec --node 26.12.1 -- vp run --filter @t3tools/client-runtime --filter @t3tools/web typecheck: passed.
• Mutation check on the pre-merge branch: removing the fallback made the hover-card test lose My GPT · high. Upstream has since deleted that browser test file.
• Not checked: browser tests (Playwright Chromium isn't installed on this machine) and an integrated pass in a real client after the merge.

These screenshots predate the upstream merge:

Before: the hover card shows the child's model only	After: it shows the model and the effort Droid reported

Built with Claude Opus 5.5 orchestrating GPT-6 Luna workers through Factory Droid in T3 Code. Upstream preparation by GPT-6 Luna, orchestrated by Claude Opus 5.5, through Factory Droid in T3 Code.