🤖 refactor: auto-cleanup

[mux-bot]

Summary

Long-lived auto-cleanup PR maintained by the Auto-Cleanup Agent. Each pass lands at most one extremely low-risk, behavior-preserving cleanup drawn from recently merged main activity, then advances the checkpoint below.

This pass

Deduplicated the repeated "Workflow metadata.argsSchema.properties must be an object" error string in src/node/services/workflows/workflowArgs.ts (added in #3575, "streamline workflow authoring") into a module-level ARGS_SCHEMA_PROPERTIES_ERROR_MESSAGE constant. The literal appeared verbatim in both assertObjectSchema and propertySchemas — the two functions that validate argsSchema.properties.

• Behavior-preserving: the constant holds the exact string literal it replaced; the isPlainObject(...) guards and throw sites are otherwise unchanged.
• Why it helps: a single source of truth keeps the two properties validators from drifting to different messages for the same failure mode.
• Scoped: single file, no import changes, no exported-API changes.

Validation

• bun test src/node/services/workflows/WorkflowService.test.ts — 57 pass / 0 fail.
• make static-check — eslint, tsc (both configs), Prettier, and the generated workflow-runtime bundle drift check pass. The only failing step is fmt-shell-check, which requires shfmt (not installed in this environment); no shell scripts were touched, so CI (which has shfmt) is unaffected.

Prior passes

• Deduplicated the four mocked-module specifier strings in src/browser/hooks/useModelsFromSettings.test.ts into shared local constants so installUseModelsModuleMocks() / restoreUseModelsModuleMocks() can't drift (after 🤖 tests: isolate renderer test globals #3574).
• Removed a redundant toggleWorkflowExpanded wrapper in src/browser/features/Tools/WorkflowRunToolCall.tsx, left behind by 🤖 fix: auto-collapse workflow tool cards #3571. The wrapper had collapsed to a pure pass-through (() => { toggleExpanded(); }); it's now inlined at its single <ToolHeader onClick={...}> call site by passing the hook's toggleExpanded directly.
• Deduplicated two repeated fallback error strings in src/browser/components/AutomationModal/AutomationModal.tsx (after 🤖 feat: centralize workflow automations #3560) into SAVE_AUTOMATION_ERROR_MESSAGE / REMOVE_AUTOMATION_ERROR_MESSAGE module-level constants. Each literal previously appeared three times across the save/remove flows; behavior-preserving since each constant holds the exact same inlined string.
• Replaced a duplicated "wfr_" literal in src/node/services/workflows/WorkflowRunner.ts with the canonical isWorkflowRunTaskId() predicate from src/node/services/tools/taskId.ts (after 🤖 feat: add first-class nested workflows #3565). Behavior-preserving inside the Array.find() predicate (both forms are falsy for a missing taskId and perform the same startsWith check for a present one).
• Removed a duplicated "chat.jsonl" literal in src/node/services/analytics/etl.ts by re-exporting CHAT_FILE_NAME from the canonical src/common/constants/paths.ts (after 🤖 perf: rotate sealed chat history out of chat.jsonl at compaction boundaries #3541).
• Extracted a harvestRecordTime helper in src/node/services/memoryConsolidationService.ts so findNewestHarvestRecord and pruneHarvestRecords derive the completedAt ?? startedAt ordering key from one place (after 🤖 feat: add memory harvest pipeline #3558).

Auto-cleanup checkpoint: 7e3277c

---

Generated with mux • Model: anthropic:claude-opus-4-8 • Thinking: xhigh • Cost: n/a

[mux-bot]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

⚠️ Auto-fixup analysis: the CI failure was a flaky test, not caused by this PR's changes. No fix pushed.

Failure: Run 27552085157 (commit 141167eb8) failed only in Test / Unit, on:
built-in deep-research workflow > keeps prototype-key source identifiers during dedupe
with a Bun engine-level crash: RuntimeError: Unreachable code should not be executed.

Why it's flaky / unrelated:

• This PR's only diff is a pure refactor of src/node/services/memoryConsolidationService.ts (extracting a harvestRecordTime helper). It has no connection to the deep-research workflow source-dedupe path that crashed.
• RuntimeError: Unreachable code should not be executed is an intermittent Bun JIT/engine crash, not a deterministic assertion failure.
• Reproduced locally on the PR head: the originally-failing test passed in 3/3 full-file runs, while a different test (auto-fix checkpoint retry preserves completed patch after HEAD advances) flaked once — confirming non-determinism in this file.
• memoryConsolidationService.test.ts: 36/36 pass.
• A subsequent CI run on the current head a913e8d48 (run 27552691244) already completed successfully.

Action: None required for this PR. The flakiness lives in builtInWorkflowDefinitions.test.ts and should be tracked/stabilized separately.

---

Generated with mux • Model: anthropic:claude-opus-4-8 • Thinking: xhigh

[mux-bot]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

Auto-cleanup pass note: the one failing check (Test / Unit) is pre-existing on main, not caused by this change.

• This pass only touches src/node/services/analytics/etl.ts (re-export the canonical CHAT_FILE_NAME instead of redefining the "chat.jsonl" literal). historyService.ts/historyService.test.ts are byte-identical to origin/main.
• The failing tests are HistoryService > getMessagesForCompactionEpoch > … (introduced by 🤖 perf: rotate sealed chat history out of chat.jsonl at compaction boundaries #3541). Test / Unit is currently red on main HEAD (a6d11e3b5) and the prior commit (a8f811f96) independently of this PR.
• All other checks pass here, including CI Static Checks (full make static-check incl. shfmt), Test / Integration, builds, E2E, Storybook, and Chromatic.

Per the auto-cleanup agent's "extremely low-risk, behavior-preserving" constraint, this pass does not attempt to fix the unrelated HistoryService test regression. Pausing for human direction on that pre-existing failure.

[mux-bot]

⚠️ Auto-fixup could not resolve CI failure: the failing tests are not caused by the cleanup commits — they are pre-existing failures already red on main. Manual intervention needed.

What failed

Test / Unit failed with 10 tests (failed run 27564995706):

• HistoryService > getMessagesForCompactionEpoch (3 tests)
• MemoryConsolidationService (7 tests, incl. 2 timeouts)

Why it is not from this branch

The same 10 tests fail on main HEAD's own CI (acb98ed6c, run 27564747006), and I reproduced them locally at the merge base a6d11e3b5 — neither of which contains this branch's two commits. Both cleanup commits are behavior-preserving and untouched by the affected code:

• 426bc8d41 only refactors memoryConsolidationService.ts ordering (identical values).
• 2df82f753 only re-exports CHAT_FILE_NAME in analytics/etl.ts.

Root cause (pre-existing, from #3558)

src/node/services/historyService.ts getMessagesForCompactionEpoch calls this.iterateForward(workspaceId, …), but iterateForward(filePath, …) expects a file path (every other call site passes getChatHistoryPath(...) / getChatArchivePath(...)). With a bare workspaceId, fs.stat hits ENOENT, the visitor never runs, the summary is never found, and the call returns Err("Compaction summary not found"). The MemoryConsolidationService harvest tests fail/time out downstream because the memory harvest reads compaction-epoch messages through this broken path. The line was introduced by 5584b6437 (#3558), which is on main.

Action

Per the auto-fixup scope ("only fix what the cleanup broke; do not add unrelated changes"), I did not push a fix. The fix belongs on main (e.g., pass this.getChatHistoryPath(workspaceId) and iterate the sealed archive too). Once main is green, this PR's CI should pass on rebase/merge.

[mux-bot]

@codex review

This pass replaces a duplicated "wfr_" literal in WorkflowRunner.ts's nested-step drift check with the canonical isWorkflowRunTaskId() predicate from taskId.ts. Behavior-preserving (the result is only used as a boolean inside Array.find(); both forms are falsy for a missing taskId).

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

@codex review

Auto-cleanup pass: removed a redundant toggleWorkflowExpanded wrapper in WorkflowRunToolCall.tsx that became a pure pass-through to the stable toggleExpanded callback after #3571. Behavior-preserving, single-file. Checkpoint advanced to f41a9a4e54e56af4369f837bc89fe9afe9224681.

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[mux-bot]

Auto-cleanup pass (checkpoint advanced to ec3314ebf, after #3574):

• Change: deduplicated the four mocked-module specifier strings in src/browser/hooks/useModelsFromSettings.test.ts into shared *_MODULE constants used by both the install and restore helpers (string→const, behavior-preserving).
• Green: Static Checks, Test / Unit (runs the edited file), Test / Windows, Test / Storybook, Test / E2E (linux + macos), Build / Linux, Smoke / Server, Visual Regression, UI Tests/Review, codecov/patch.
• Red — unrelated/flaky: Test / Integration failed on live-AI/streaming/runtime cases (tests/ipc/runtime/runtimeExecuteBash stream timeouts, tests/ipc/streaming/sendMessage.* model-call assertions, tests/ipc/config/mcpConfig). These have no code path to useModelsFromSettings; the unit gate that exercises my edit is green. The integration token here can't re-run the job — please re-run Test / Integration before merge.
• Codex: the connector is not configured in this environment (each @codex review returns a "connect your account" auto-reply), so the Codex gate can't be cleared here.

[mux-bot]

⚠️ Auto-fixup determined this CI failure is a flaky live-API integration test, not a regression from the cleanup commit — so no code was pushed.

Failed job: Test / Integration (all other jobs — Static Checks, Unit, Storybook, E2E linux/macos, Windows, builds — passed).

Failures (all stream timeouts against the real model API):

• tests/ipc/runtime/runtimeExecuteBash.test.ts — 4 cases: Stream timeout after 15000ms waiting for stream-end
• tests/ipc/config/mcpConfig.test.ts — 1 case: Stream did not emit stream-end event

These tests send real messages to a model (HAIKU) and wait for stream-end; the timeouts indicate API latency/rate-limiting, not a code defect.

Why it's not caused by this PR: the cleanup diff only touches UI (AutomationModal.tsx, WorkflowRunToolCall.tsx), a test mock dedupe (useModelsFromSettings.test.ts), and three behavior-preserving node refactors (etl.ts re-export of CHAT_FILE_NAME, an extracted harvestRecordTime ranking helper in memoryConsolidationService.ts, and a wfr_-prefix helper swap in WorkflowRunner.ts). None of these are in the agent streaming loop, runtime bash tool, MCP integration, or IPC stream helpers exercised by the failing tests.

Recommended action: re-run the failed Test / Integration job. If it consistently fails on re-run, manual investigation of the integration test infra (API availability/timeouts) is needed.

[mux-bot]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.