feat(evals v4): OTEL agent spans for claude_code/codex harnesses

[miguelg719]

Why

Stack 5/6. External harnesses drive their own SDKs (claude-agent-sdk/codex-sdk) which emit no OTEL.

What

• Wrap runClaudeCodeAgent/runCodexAgent in withHarnessAgentSpan so their runs nest an agent node under the eval task-root span — parity with the stagehand harness (whose native RPC spans flow in via the global provider).
• No-op when EVAL_TRACE_TRANSPORT is unset.

Testing

Validated live: a claude_code run on Browserbase produced an 85-span LangSmith trace (task-root → agent.claude_code → native page/act/observe spans).

Base: #2760

---

Summary by cubic

Adds OpenTelemetry agent spans for the claude_code and codex harnesses so eval traces show a nested agent span under the task root. Previously these harnesses emitted no OTEL; now each run sets harness, model, task, and instruction span attributes without changing execution.

• Wraps runClaudeCodeAgent/runCodexAgent with withHarnessAgentSpan in packages/evals/framework/benchHarness.ts.
• No-op unless EVAL_TRACE_TRANSPORT is set; behavior matches the stagehand harness; no migration required.

^{Written for commit db1f171. Summary will update on new commits.}

[changeset-bot]

⚠️ No Changeset found

Latest commit: db1f171

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[cubic-dev-ai]

All reported issues were addressed across 1 file

Architecture diagram

sequenceDiagram
    participant Eval as Eval Runner
    participant Cli as claudeCodeHarness/codexHarness
    participant Span as withHarnessAgentSpan
    participant OTEL as OTEL/LangSmith
    participant Agent as claude-code/codex SDK

    Note over Eval,Agent: Runtime flow for external harnesses

    Eval->>Cli: execute plan
    Cli->>Cli: Prepare tool adapter, plan, signal, verifier

    alt EVAL_TRACE_TRANSPORT is set
        Cli->>Span: Wrap agent execution
        Span->>Span: Create agent span (attributes: harness, model, task, instruction)
        Span->>OTEL: Start agent span
        Span->>Agent: Execute runClaudeCodeAgent/runCodexAgent
        Agent-->>Span: Agent result
        Span->>OTEL: End agent span (attach to eval task-root)
        Span-->>Cli: Return result
    else EVAL_TRACE_TRANSPORT is unset
        Cli->>Agent: Execute runClaudeCodeAgent/runCodexAgent (no-op span)
        Agent-->>Cli: Agent result
    end

    Cli->>Cli: Run toolAdapter cleanup and deregister carrier
    Cli-->>Eval: Return task result

    Note over OTEL,Agent: Native page/act/observe spans from stagehand flow in via global provider

Loading

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>