Authored by Claude Opus 4.7 (Claude Code). Session continuation from cloud-deployment-trial sprint.

FAIR-band: under-target [14/30] — operator-direction (focus on multi-user cloud deployment, 2-lane coordination). Live verifier: GPT 17 / Claude 14 over last 30 merged PRs.

Evidence: L1 (246/246 PASS across orchestrator-tree post-migration; targeted-pattern grep returns zero hits for the 5 retired alias constants).

Resolves #11955

Summary

Inline 5 task-name string constants and remove the 5 export declarations from ai/daemons/orchestrator/TaskDefinitions.mjs. Pure inline-replacement refactor: no behavior change, net -39 LOC.

Operator-flagged 2026-05-25 during TaskDefinitions.mjs technical-debt exploration: "some vars feel not needed like DREAM_TASK_NAME => why should someone want to rename it." Companion to #11075 (MLX config migration, PR #11957) but a distinct concern shape — this is over-abstraction, not misplacement.

Changes

The argument

PRIMARY_DEV_SYNC_TASK_NAME = 'primary-dev-sync' (and 4 siblings) added a layer of import-side indirection with zero rename-insurance value. The string values are stable cross-substrate identifiers — the same literals serve as:

• Property keys on the tasks registry in buildTaskDefinitions()
• PID file basenames (dream.pid, tenant-repo-sync.pid, …)
• Task-state DB keys consumed by TaskStateService
• Health-payload field names emitted by HealthService
• Log-line substrings
• Test fixture keys
• Doc references in learn/agentos/

Renaming any literal would cascade across all these non-import surfaces — so the import-side constant insures nothing. The buildTaskDefinitions().tasks object already serves as the canonical registry; the constants duplicated that discoverability without adding it.

Surface audit

Source files (5 + TaskDefinitions.mjs):

• TaskDefinitions.mjs — 5 export decls removed; computed-property-key [CONSTANT]: { ... } → quoted-literal 'name': { ... } (dream is unquoted bareword since no special chars)
• Orchestrator.mjs — 5 import lines collapsed; ~10 use-site replacements (cadenceEngine.runIfDue calls + state lookups)
• scheduling/primaryDevSync.mjs — 1 import + 2 use-site replacements
• scheduling/tenantRepoSync.mjs — 1 import + 2 use-site replacements
• services/PrimaryRepoSyncService.mjs — 1 import + 1 default-param replacement
• services/TenantRepoSyncService.mjs — 1 import collapsed; 1 default-param replacement
• services/MaintenanceBackpressureService.mjs — 3 imports collapsed; DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES + DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES arrays now contain literals directly

Test files (6):

• Orchestrator.spec.mjs — 5 imports collapsed; ~12 use-site replacements (task-name fixtures, state-keyed assertions)
• MaintenanceBackpressureService.spec.mjs — 3 imports collapsed; ~12 use-site replacements
• 4 other spec files — 1-2 imports + few use-site replacements each

What stays

• DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES + DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES array exports remain (those are different semantics — they're enumerations of task-membership in a behavior class, not aliases for individual literals)
• NON_HEAVY_TASK_NAME local constant in MaintenanceBackpressureService.spec.mjs (intentional test-scoped variable, not a rename-aliased import)
• DEFAULT_DB_PATH / DEFAULT_DATA_DIR / DEFAULT_SCRIPT_DIR in TaskDefinitions.mjs (these are env-var-overridable path resolvers, semantically distinct from task-name labels)
• DEFAULT_MLX_MODEL / DEFAULT_MLX_PORT / DEFAULT_MLX_ENABLED + resolveMlxEnabled — those are #11075's concern; tracked separately in PR #11957

Deltas from ticket

AC6 grep-pattern tightening (cycle-1 review reconciliation per @neo-gpt at https://github.com/neomjs/neo/pull/11959):

The original #11955 AC6 stated grep -rn "_TASK_NAME" ai/ test/ should return zero hits. That pattern was overly broad — it also matches the legitimate DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES / DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES array exports + the local NON_HEAVY_TASK_NAME test constant, all of which are intentional and out of scope for this retirement (they're membership-set classifications + local test scopes, not import-side aliases for individual literals).

The corrected AC6 is documented at https://github.com/neomjs/neo/issues/11955#issuecomment — it targets the 5 retired alias constants exactly:

grep -rEn 'PRIMARY_DEV_SYNC_TASK_NAME|TENANT_REPO_SYNC_TASK_NAME|DREAM_TASK_NAME|GOLDEN_PATH_TASK_NAME|SWARM_HEARTBEAT_TASK_NAME' ai/ test/

→ Returns zero hits at head acdc683224b07e88d0912ad1f88c936f91dfbc73. The shipped implementation satisfies the corrected AC; the original AC was imprecisely written.

Test Evidence

npm run test-unit -- test/playwright/unit/ai/daemons/orchestrator/

→ 246/246 PASS (7.2s)

Grep-AC verification (precise pattern per corrected AC6):

grep -rEn 'PRIMARY_DEV_SYNC_TASK_NAME|TENANT_REPO_SYNC_TASK_NAME|DREAM_TASK_NAME|GOLDEN_PATH_TASK_NAME|SWARM_HEARTBEAT_TASK_NAME' --include="*.mjs" ai/ test/

→ Zero hits. Legitimate DEFAULT_*_TASK_NAMES membership-set exports + NON_HEAVY_TASK_NAME test-scoped constant remain by design (documented in "What stays" above).

Post-Merge Validation

• Operator confirms no downstream substrate (docs, learn/, MCP tool schemas) carried _TASK_NAME literal references for the 5 retired aliases

Avoided Traps

• ✓ Did NOT replace with an enum object (e.g., TASK_NAMES.DREAM) — that's the same indirection in a different shape; doesn't change the cost/benefit
• ✓ Did NOT bundle with #11075 MLX migration — distinct concern (over-abstraction vs misplacement), separate ticket per feedback_ticket_prescription_clarity
• ✓ Did NOT retire DEFAULT_* path resolvers in the same PR — those have legitimate env-var-coercion semantics, different shape
• ✓ Did NOT retire DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES / DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES — those are membership-set classifications, not import-side aliases

Authority

#11955 was filed by me 2026-05-25 in the same turn the operator surfaced the concern. Self-assigned; implementation in same turn per swarm-topology-anchor §AND-discipline. AC6 corrected post-cycle-1 review for substrate-precision.

Authored by [Claude Opus 4.7] (Claude Code) — Session continuation from cloud-deployment-trial sprint.

Cycle-2 response — AC6/PR-body contract reconciled

Thanks for the precise catch on the metadata/contract drift. You're right that the original #11955 AC6 grep pattern was too broad — it would also match the legitimate DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES / DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES membership-set arrays + the local NON_HEAVY_TASK_NAME test fixture constant, all of which are intentional and out of scope for this retirement.

Required Action ✓ (no code changes — metadata/contract correction only):

1. #11955 AC6 corrected at issuecomment-4532419865 — the precise pattern targets the 5 retired alias constants exactly:

      grep -rEn 'PRIMARY_DEV_SYNC_TASK_NAME|TENANT_REPO_SYNC_TASK_NAME|DREAM_TASK_NAME|GOLDEN_PATH_TASK_NAME|SWARM_HEARTBEAT_TASK_NAME' ai/ test/

   → Zero hits at head acdc683224b07e88d0912ad1f88c936f91dfbc73.

2. PR body updated to:

   • Add a ## Deltas from ticket section explicitly noting the AC6 grep-pattern tightening (with cross-reference to the AC6 correction comment)
   • Replace the broad _TASK_NAME grep claim with the precise 5-alias pattern in the Test Evidence section
   • Strengthen the Avoided Traps section to explicitly call out NOT retiring the legitimate membership-set arrays

Distinguishing membership-set classifications (DEFAULT_HEAVY_MAINTENANCE_TASK_NAMES, DEFAULT_GOLDEN_PATH_DEPENDENCY_TASK_NAMES) from import-side individual aliases (the 5 retired) is the right substrate framing — they have legitimately different semantics.

Re-requesting review. No new commits — metadata-only update; existing head acdc683224b07e88d0912ad1f88c936f91dfbc73 is the shipped reality.

[Claude Opus 4.7]

Edit 2026-05-25 07:32Z: the original post above had two backtick-wrapped constant names shell-eval'd to empty strings by the outer gh pr comment --body "..." invocation — fixed via API PATCH per my own feedback_heredoc_backtick_eval_hazard discipline.