refactor(tests): build the workflow agents in the error-event tests from config

[kalenkevich]

Link to Issue or Description of Change

Problem:

Two spots in node_error_event_test.ts built a workflow agent as:

new WorkflowAgent(new Workflow({name: 'wf', edges: [...]}))

WorkflowAgent has a convenience overload that takes the workflow config directly and constructs the Workflow itself, so the inner constructor was only adding a nesting level.

Solution:

Use the config overload in both places.

The other two new WorkflowAgent(existingWorkflow) uses in the suite are deliberate and stay — workflow_agent_test.ts covers that overload specifically ("still accepts a pre-built Workflow, with optional overrides").

No behaviour change: the config form constructs the same Workflow from the same config, including the rerunOnResume default.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

No new tests — this is a readability change inside existing ones.

npm run ts:check          clean
npx vitest --project unit:core    212 files, 2963 passed

Manual End-to-End (E2E) Tests:

Not applicable.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

Touches only a file from #657, so it is independent of the other workflow PRs in flight.

For the record on the larger question this came out of — whether the sample and integration tests could drop WorkflowAgent now that #680 lets a Workflow be a Runner root — they cannot, and it is worth knowing why. agent_loader.ts:277 selects exports with isBaseAgent, so a sample exporting a bare Workflow would not load at all; and agent_graph.ts:50 reads rootAgent.workflow through isGraphWorkflowAgent to render the graph added in #654. Widening those two seams is the same work as dissolving WorkflowAgent, which belongs with the LLMAgentWrapper cleanup rather than being done piecemeal.

[AmaadMartin]

Reviewed at 686c52d. The refactor is correct. WorkflowAgent has a constructor(config: WorkflowConfig) overload that builds new Workflow(config) internally (workflow_agent.ts), so the config form produces the same Workflow as the old explicit wrap. No behaviour change. Workflow stays imported and used across the file. No any, no suppressions, no instanceof added.

Not approving yet: run-tests (windows-latest) is still pending. macOS and ubuntu passed. Approve once Windows is green.

[AmaadMartin]

Approving at 686c52d. My earlier review said the refactor was correct and then withheld approval because run-tests (windows-latest) was pending. That was wrong, and it was the reviewer's fault rather than anything about this PR.

The cause is worth naming: the review skill still said to treat a pending check as not green and wait for it. An unattended reviewer cannot wait — it only re-reviews when the head SHA changes, and CI turning green does not change the SHA — so "approve once Windows is green" was a promise it could never keep. The skill has been corrected.

On the code, unchanged from my earlier read: WorkflowAgent's constructor(config: WorkflowConfig) overload builds new Workflow(config) internally, so the config form produces the same Workflow as the old explicit wrap. No behaviour change, Workflow stays imported and used, and the diff adds no suppressions and no instanceof. All checks are green now in any case. LGTM.