Stub getCurrentAuthedUser/getUserOrganizations on NullBaseOctoKitService

[meganrogge]

Problem

When running an msbench eval against sessionTarget: agent-host-copilot (e.g. runtest local --benchmark vscbench.say_hello --config doc/examples/configs/agent-host.vscode.agent.yaml), the Copilot extension host logs:

WARN this.octoKitService.getCurrentAuthedUser is not a function

…and the agent-host benchmark hangs.

Root cause

In src/extension/extension/vscode/services.ts:

builder.define(IOctoKitService, isScenarioAutomation
    ? new SyncDescriptor(NullBaseOctoKitService)
    : new SyncDescriptor(OctoKitService));

The eval container sets IS_SCENARIO_AUTOMATION=1, so NullBaseOctoKitService is injected as IOctoKitService. But NullBaseOctoKitService extends BaseOctoKitService, which only implements the *WithToken variants (getCurrentAuthedUserWithToken, getUserOrganizationsWithToken). The no-arg interface methods getCurrentAuthedUser() / getUserOrganizations() are only implemented on the real OctoKitService (which has access to IAuthenticationService to fetch the token).

Why agent-host specifically

The only non-test caller of IOctoKitService.getCurrentAuthedUser is GitHubOrgChatResourcesService.computePreferredOrganizationName (src/extension/agents/vscode-node/githubOrgChatResourcesService.ts). That service is consumed by GitHubOrgCustomAgentProvider and GitHubOrgInstructionsProvider — features that only the agent-host code path activates to discover GitHub-org-backed custom agents and instruction files. Regular chat / inline chat / edit mode never traverse this service, which is why every other sessionTarget runs cleanly through the eval and only agent-host-copilot triggers it.

Fix

Add stub overrides on NullBaseOctoKitService:

• getCurrentAuthedUser() → returns the NullUser placeholder (consistent with the existing getCurrentAuthedUserWithToken stub).
• getUserOrganizations() → returns [].

The caller already handles the empty / unsigned-in case correctly (it just skips org discovery), so this unblocks the agent-host eval path without changing real-user behavior.

Validation

Type-checks cleanly. End-to-end validation requires this fix to ship in Insiders so the eval Docker image picks it up; once that lands, re-run runtest local --benchmark vscbench.say_hello --config doc/examples/configs/agent-host.vscode.agent.yaml.

Companion to #314526 (which fixes the parallel agentHostPermissionService workbench-DI error encountered on the same eval run).

[Copilot]

Pull request overview

This PR updates the Copilot extension’s GitHub Octokit “null” implementation used in scenario automation so that agent-host org discovery code paths don’t hit missing-method runtime errors during MSBench eval runs.

Changes:

• Add a stub getCurrentAuthedUser() implementation to NullBaseOctoKitService.
• Add a stub getUserOrganizations() implementation to NullBaseOctoKitService.

Show a summary per file

File	Description
extensions/copilot/src/platform/github/common/nullOctokitServiceImpl.ts	Adds missing no-auth stubs on the null Octokit service to prevent agent-host callers from failing at runtime.

Copilot's findings

• Files reviewed: 1/1 changed files
• Comments generated: 2

[github-actions]

Base: b6cb647b Current: 01e9f170

No screenshot changes.

[meganrogge]

/requires-eval-assessment terminalbench2 gpt-5.4,claude-opus-4.6,claude-opus-4.7

vscodeSettings:
  chat.tools.compressOutput.enabled: true

[vs-code-engineering]

⏳ Queued vscode build for 2a5bc290f3c028feb2e774c4e7239bbd13a2720d (step 1/2).

• Build: https://dev.azure.com/monacotools/Monaco/_build/results?buildId=436745
• When this succeeds, the eval-assessment publish build will be queued automatically.

[vs-code-engineering]

🚀 Queued eval-assessment publish build for 01e9f1709931b3baba7031e5d014ca97728675a5 (step 2/2).

• Pipeline run: https://dev.azure.com/monacotools/Monaco/_build/results?buildId=436760
• On success, publishes @vscode/vscode-copilot-evaluation-agent@0.0.0-dev.01e9f17099 on the dev tag.

[vs-code-engineering]

🔬 Queued eval-assessment benchmark for 51555b0058.

• Package: @vscode/vscode-copilot-evaluation-agent@0.0.0-dev.01e9f17099 (dev tag)
• Benchmark: terminalbench2
• VS Code overrides: applied from PR comment (2 line(s))
• Model: claude-sonnet-4.5
• Tracking issue: https://github.com/github/evald/issues/20622

Results will be posted back here when the run completes.

[vs-code-engineering]

✅ Eval-assessment build published.

• Package: @vscode/vscode-copilot-evaluation-agent@0.0.0-dev.01e9f17099 (tag: dev)
• Install: npm install @vscode/vscode-copilot-evaluation-agent@0.0.0-dev.01e9f17099
• Pipeline run: https://dev.azure.com/monacotools/Monaco/_build/results?buildId=436760