Summary

• extract the codex auth check entry point into a dedicated command module
• add focused unit coverage for the command wrapper while preserving the existing health-check behavior

What Changed

• added lib/codex-manager/commands/check.ts as a dedicated check command wrapper
• updated lib/codex-manager.ts so the check dispatch delegates to the new command module
• added test/codex-manager-check-command.test.ts to verify the command always invokes runHealthCheck({ liveProbe: true })

Validation

• npm run lint
• npm run typecheck
• npm run test -- test/codex-manager-check-command.test.ts test/codex-manager-status-command.test.ts test/codex-manager-switch-command.test.ts test/codex-manager-cli.test.ts
• npm run build

Risk and Rollback

• Risk level: low
• Rollback plan: revert c37beba to restore the inline check dispatch

Additional Notes

• this PR stays stacked on #151 so the command split remains easy to review in one-worktree slices

note: greptile review for oc-chatgpt-multi-auth. cite files like lib/foo.ts:123. confirm regression tests + windows concurrency/token redaction coverage.

Greptile Summary

extracts the codex auth check dispatch into lib/codex-manager/commands/check.ts using the same dependency-injection pattern established by the sibling status and switch command modules already landed in #151.

• runCheckCommand is an 8-line wrapper that injects runHealthCheck as a dep — clean, testable, no behavior change
• codex-manager.ts dispatch is now a single-line delegation: return runCheckCommand({ runHealthCheck })
• unit tests cover the success path (returns 0) and the rejection path (propagates error), consistent with sibling test files
• lib/AGENTS.md is not updated to document the commands/ subdirectory; ai agents using the knowledge base will miss check.ts, status.ts, and switch.ts — worth a follow-up update
• no windows filesystem, token safety, or concurrency concerns in this change — runHealthCheck is a single sequential call with no shared mutable state

Confidence Score: 5/5

• safe to merge — pure refactor with no behavior change and existing integration coverage in codex-manager-cli.test.ts
• the dispatch change is semantically identical to the original inline call, the new module follows established patterns, and the integration tests in codex-manager-cli.test.ts already exercise the check command end-to-end. only gap is a documentation update to AGENTS.md and a minor test contract assertion.
• lib/AGENTS.md — needs updating to document the new commands/ subdirectory

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant CLI as runCodexMultiAuthCli
    participant CMD as runCheckCommand
    participant HC as runHealthCheck

    CLI->>CMD: runCheckCommand({ runHealthCheck })
    CMD->>HC: runHealthCheck({ liveProbe: true })
    HC-->>CMD: resolves (void)
    CMD-->>CLI: return 0

This is a comment left during a code review.
Path: lib/AGENTS.md
Line: 33

Comment:
**`commands/` subdirectory not documented in knowledge base**

`lib/AGENTS.md` still shows `codex-manager/` as containing only `settings-hub.ts`. this PR (and its base #151) add a `commands/` subdirectory (`check.ts`, `status.ts`, `switch.ts`). any ai agent referencing this guide will miss the new modules entirely.

```suggestion
│   ├── codex-manager/    # settings-hub extraction (2100 lines)
```
should be updated to:
```
│   ├── codex-manager/    # settings-hub + extracted command modules
│   │   ├── commands/     # check, status, switch command wrappers
│   │   └── settings-hub.ts  # extracted interactive settings (2100 lines)
```
similarly, the `WHERE TO LOOK` and `STRUCTURE` sections in `lib/AGENTS.md` need a row for the `commands/` subdir.

**Context Used:** lib/AGENTS.md ([source](https://app.greptile.com/review/custom-context?memory=ffbed5b7-f299-4c32-b3f9-f7c095ca8632))

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: test/codex-manager-check-command.test.ts
Line: 1-32

Comment:
**missing non-throw failure path test**

the two tests cover `return 0` and `rejects`. there is no test covering the case where `runHealthCheck` resolves successfully but an in-process signal (e.g. a console-printed error summary) means the check conceptually "failed" with exit 0 — which is the current behavior. this is intentional since `runHealthCheck` returns `void`, but it means callers cannot distinguish a clean "all accounts healthy" from a "some accounts degraded but no throw". worth asserting explicitly that the command unconditionally returns `0` on any resolve, to lock the contract and document the limitation.

How can I resolve this? If you propose a fix, please make it concise.

_{Last reviewed commit: "test(check): cover w..."}

Context used:

• Context used - lib/AGENTS.md (source)