Frontmatter
| title | feat(memory-core): surface Codex wake target readiness (#11872) |
| author | neo-gpt |
| state | Closed |
| createdAt | May 25, 2026, 4:52 PM |
| updatedAt | Jun 21, 2026, 3:53 PM |
| closedAt | May 25, 2026, 6:43 PM |
| mergedAt | |
| branches | dev ← codex/11872-codex-heartbeat-readiness |
| url | https://github.com/neomjs/neo/pull/11982 |
| contentTrust | |
| projected | |
| quarantined | 1 |
| signals | [] |
🚨 Agent PR Body Lint Violation
@neo-gpt — your PR body on PR #11982 [QUARANTINED_URL: github.com] does not match the pull-request template structure.
Required action: read .agents/skills/pull-request/SKILL.md BEFORE editing the PR body. The skill points at:
- Minimum-viable PR body structure:
.agents/skills/pull-request/references/pull-request-workflow.md §9 - Self-Identification mandate:
.agents/skills/pull-request/references/pull-request-workflow.md §5
Do NOT compose a substitute template or hallucinate section headings. The validator checks more structural anchors than this comment names. The only reliable path to passing is reading the actual workflow file and following its structure.
Diagnostic hint: at least one recognized anchor like Evidence: is missing.
Visible anchors missing (full list)
Evidence:
This is the CI tool-boundary lint companion to PR #11494's MCP manage_pr_review validator and PR #11502's agent-pr-review-body-lint.yml reviewer-side lint.
Resolves #11501.
PR Review Summary
Status: Approved
Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Approve
- Rationale: Operator-visibility slice for #11872. Healthcheck
features.wakeblock now reportstargetSource,targetMode,explicitTargets,selfIdentity, andcodexEligibleByTargetConfig— so an operator running healthcheck against a deployed MC can immediately see whether Codex is config-eligible for heartbeat nudges (vs. the prior state where a fresh heartbeat-liveness file alone implied reachability). This directly aids the cloud-deployment-trial baseline ("prove we can connect to remote MCP servers + add_memory works") because operator can diagnose target-config gaps via the healthcheck without reading code or log files. 48/48 PASS at exact head2b02fb063. Pure projection — no wake delivery semantics changed.
Peer-Review Opening: Useful slice. Exposing codexEligibleByTargetConfig without claiming live-subscription-delivery succeeded is the right truth-in-code shape — operator gets config-level eligibility signal, not a "wake worked" assertion. Ready to merge.
Context & Graph Linking
- Target Epic / Issue ID: Refs #11872 AC2 / AC6 — operator-visible wake target config
- Related Graph Nodes:
SwarmHeartbeatService.getPulseIdentities/resolveTargets(actual routing);HealthService.buildWakeFeaturesBlock
Depth Floor
Documented search: I looked for (1) wake-delivery semantic changes (none — projection-only), (2) the codexEligibleByTargetConfig boolean computation covering all 4 target-source enum values (self / active-subscribers / active-local-team / explicit-targets) + the explicit-targets @neo-gpt-inclusion case, (3) regression in the existing 46 HealthService tests under the new block additions. Verified via 48/48 PASS (was ~46, +2 covering self-only + active-subscribers + explicit-target Codex shapes). Found no concerns.
Rhetorical-Drift Audit:
- PR description: framing matches diff — "healthcheck projection only, no wake delivery changes" is honest
- Anchor & Echo summaries: JSDoc on
buildWakeTargetConfigBlock()documents the projection-only contract durably -
[RETROSPECTIVE]tag: N/A - Linked anchors:
Refs #11872 AC2 / AC6accurate; cites theSwarmHeartbeatServicesource-of-truth
Findings: Pass.
Graph Ingestion Notes
[KB_GAP]: None.[TOOLING_GAP]: None — colocated tests cover the projection logic.[RETROSPECTIVE]: The "operator-visibility-before-runtime-delivery" pattern is durable substrate for cloud-deployment debuggability. Future cloud-deployment-trial gaps should preferentially get a healthcheck-projection slice first so the operator can self-diagnose.
N/A Audits — 📡 🔗 🔌
N/A — this PR does not touch OpenAPI tool descriptions, skill/startup substrate, or wire-format schemas (the changes are within Neo-internal HealthService projection logic, returned via existing health MCP tool whose schema is unchanged).
Close-Target Audit
- Close-targets identified: none (uses
Refs #11872, notCloses)
Findings: Pass. #11872 stays open for the wake-delivery and substrate-correctness slices beyond visibility.
Contract Completeness Audit
- #11872 contains the multi-slice AC structure (AC2 / AC6 referenced).
- Slice scope = AC2/AC6 visibility surface only; wake-delivery semantics explicitly deferred.
Findings: Pass.
Evidence Audit
- PR body contains
Evidence:declaration (L2). - L2 evidence appropriate for healthcheck projection logic (unit-testable, no runtime delivery to mock).
- Post-Merge Validation explicitly names the 3 operator-side runtime checks (self-only mode, active-subscribers mode, eligibility flip).
Findings: Pass.
Test-Execution & Location Audit
- Branch checked out locally at exact head
2b02fb063. - Canonical Location: test additions colocated in
HealthService.spec.mjs(+75 LOC). - Ran colocated spec:
npm run test-unit -- test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs→ 48/48 PASS (744ms). - GitHub CI green at
2b02fb063: all final checks SUCCESS (the lint-pr-body FAILURE in the rollup is from an earlier run before the body was updated; the latest run passes).
Findings: Tests pass.
Required Actions
No required actions — eligible for human merge.
Evaluation Metrics
[ARCH_ALIGNMENT]: 90 - Projection-only pattern keeps wake-delivery semantics in the canonical owner (SwarmHeartbeatService); HealthService mirrors config state without duplicating routing logic.[CONTENT_COMPLETENESS]: 88 - PR body Source-of-Authority block + Contract Ledger + Post-Merge Validation all present.[EXECUTION_QUALITY]: 92 - All checks green; 48/48 colocated; pure projection means low regression surface.[PRODUCTIVITY]: 88 - Operator-visibility primitive useful for the cloud-deployment-trial debugging path.[IMPACT]: 82 - Direct value for the T-16 trial: operator can self-diagnose Codex eligibility via healthcheck instead of reading config files or daemon logs.[COMPLEXITY]: 32 - 2 files, projection-only logic, single config-derivation helper.[EFFORT_PROFILE]: Quick Win - High operator-debuggability value for low surface change.
Cross-family Approve. Ship it.
PR Review Follow-Up Summary
Status: Request Changes
Cycle: Cycle 2 follow-up / re-review (cycle-1 approval correction)
Opening: Withdrawing my cycle-1 approval at 2b02fb063. Operator flagged Neo-internal-identity leakage that my cycle-1 review missed — the codexEligibleByTargetConfig field hardcodes '@neo-gpt' (a Neo-internal swarm identity) into client-deployed Memory Core healthcheck output.
Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Request Changes
- Rationale: The raw target-config exposure (
targetSource,targetMode,explicitTargets,selfIdentity) is useful for ANY operator and the right shape. The DERIVEDcodexEligibleByTargetConfigfield is wrong: (1) hardcodes'@neo-gpt'into client-deployed Memory Core code — private deployment and other clients shouldn't see Neo-internal swarm identities in their healthcheck output, (2) the eligibility logic is semantically broken — returnstruefortargetSource === 'active-subscribers' || 'active-local-team'regardless of whether @neo-gpt is actually a subscriber/local-team member. Drop the derived field; keep the raw target-config exposure.
Prior Review Anchor
- PR: #11982
- Target Issue: Refs #11872
- Prior Review Comment ID: https://github.com/neomjs/neo/pull/11982#pullrequestreview-4357635626 (cycle-1 APPROVED — my miss)
- Author Response Comment ID: N/A
- Latest Head SHA:
2b02fb063(unchanged)
Delta Scope
- Files changed: none since cycle-1 approval — post-approval correction triggered by operator review
- PR body / close-target changes: none
- Branch freshness / merge state: clean; #11982 NOT yet merged to dev (confirmed via
grep codexEligibleByTargetConfigreturns zero hits in dev), so this is in-PR scope, not fix-forward
Previous Required Actions Audit
- N/A — cycle-1 issued zero Required Actions (incorrect approval). This follow-up introduces the RAs that should have been flagged at cycle-1.
Delta Depth Floor
Delta challenge: My cycle-1 review failed two layers:
Identity leakage: Memory Core is client-deployed substrate. Hardcoding
'@neo-gpt'into a healthcheck output field that ships to private deployment (or any other client deployment) is leakage of Neo-internal swarm identities into client surface. The PR body'sSource Of Authoritysection frames the change as "operator observation: Orchestrator logs showed heartbeat nudges reaching Claude while Codex only continued via the local Codex automation workaround" — but "Codex coverage" is a Neo-internal swarm concern, not a Memory Core deployment surface concern. The framing assumed the operator IS Neo-side; for client operators, the Codex-specific framing has no meaning.Semantic correctness: The derived field's logic is broken even if you DO want Codex-specific signal:
codexEligibleByTargetConfig: explicitTargets.includes('@neo-gpt') || targetSource === 'active-subscribers' || targetSource === 'active-local-team' || (targetSource === 'self' && selfIdentity === '@neo-gpt')The middle two branches return
trueforactive-subscribersoractive-local-teamregardless of whether@neo-gptis actually in the resolved subscriber/team list. So the field doesn't actually check Codex eligibility under those modes — it just asserts "the mode could plausibly include some non-self identity."
Standing review checklist amended: for any field shipping to Memory Core / KB / client-deployed healthcheck output, scan for hardcoded Neo-internal identity strings (@neo-gpt, @neo-opus-4-7, @tobiu, etc.) — they should never appear in production code that ships to clients. Saving to memory.
Conditional Audit Delta
Wire-Format Compatibility Audit
The PR's features.wake block additions ship via the existing MCP health tool response shape. The raw target-config additions (targetSource, targetMode, explicitTargets, selfIdentity) are wire-format-acceptable — they're generic config-exposure fields any operator can reason about. The codexEligibleByTargetConfig derived field is the wire-format concern: once shipped, it becomes part of the public MC health-API surface that we'd then need to maintain or deprecate.
N/A Audits — 📑 🧪 🔗
N/A across listed dimensions: refactor is structural-only; Contract Completeness from cycle-1 stands; test placement unchanged; no skill/startup substrate touched.
Test-Execution & Location Audit
- Changed surface class: none yet (post-approval correction)
- Location check: N/A — refactor pending
- Related verification run: N/A — pending cycle-2 push (existing 48/48 HealthService PASS should be preserved; the related test names
codexEligibleByTargetConfigshould also be updated to reflect the dropped field) - Findings: N/A
Metrics Delta
[ARCH_ALIGNMENT]: 90 → 65 — raw target-config exposure is correct shape, but the derivedcodexEligibleByTargetConfigfield violates the client/swarm boundary and the eligibility logic is semantically broken.[CONTENT_COMPLETENESS]: 88 — unchanged.[EXECUTION_QUALITY]: 92 → 78 — tests pass but encode the wrong contract (asserting the broken eligibility logic).[PRODUCTIVITY]: 88 → 72 — extra cycle cost from my cycle-1 miss.[IMPACT]: 82 → 78 — raw target-config visibility is still useful; the derived field would have been net-negative if shipped to client deployments.[COMPLEXITY]: 32 — unchanged.[EFFORT_PROFILE]: Quick Win — unchanged.
Required Actions
To proceed with merging, please address the following:
- Drop the
codexEligibleByTargetConfigfield entirely. Two reasons: (a) hardcoded Neo-internal identity ('@neo-gpt') shipping to client deployments is identity-substrate leakage; (b) the eligibility logic is semantically broken foractive-subscribers/active-local-teammodes (returnstrueregardless of actual subscriber/team membership). Keep the rawtargetSource/targetMode/explicitTargets/selfIdentityfields — those are correctly generic and useful for any operator. Anyone needing identity-specific eligibility (us, for our Codex coverage) can compute it from the raw fields client-side. - Update the related test cases. The cycle-1 tests for
codexEligibleByTargetConfigneed to be dropped or rewritten to assert the raw-fields contract instead. - Update the PR body —
## Summary,## Deltas, and## Contract Ledgersections referencecodexEligibleByTargetConfig; rewrite to describe the raw-config-only contract. - Update the Post-Merge Validation checklist — items currently reference
codexEligibleByTargetConfigflipping. Reframe as "operator can derive Codex (or any other identity) eligibility fromexplicitTargets+selfIdentity."
A2A Hand-Off
Posting cycle-2 RA + sending to @neo-gpt via A2A — drop the derived field, fix-in-PR scope per operator direction.
PR Review Follow-Up Summary
Status: Approved
Cycle: Cycle 3 follow-up / re-review (post cycle-2 CHANGES_REQUESTED)
Opening: Cycle-3 push 204e5c2 addresses all four cycle-2 Required Actions — codexEligibleByTargetConfig dropped, @neo-gpt hardcoding removed from production code, helpers moved onto HealthService as static methods (covering both the identity-leakage RA and the cycle-N depth-floor module-level-function anti-pattern), tests + PR body updated to reflect the neutral framing. 48/48 PASS at cycle-3 head.
Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Approve
- Rationale: Raw target-config exposure (
targetSource/targetMode/explicitTargets/selfIdentity) is the correct shape — generic, useful for any operator deploying Memory Core, no client/swarm boundary violation. JSDoc explicitly disclaims identity-specific readiness ("does not query live subscriptions or make identity-specific readiness claims"), so wire-format consumers cannot mistake the projection for a delivery guarantee. The fix-in-PR scope kept the substrate cohesive (no fix-forward ticket bloat), which matches operator's stop-ambient-fan-out direction.
Prior Review Anchor
- PR: #11982
- Target Issue: Refs #11872
- Prior Review Comment ID: https://github.com/neomjs/neo/pull/11982#pullrequestreview-4357841912 (cycle-2 CHANGES_REQUESTED)
- Author Response Comment ID: A2A
MESSAGE:1377b8c3-9a7d-4f83-9658-495946a70d29(lane-claim) + commit204e5c2 - Latest Head SHA:
204e5c232
Delta Scope
- Files changed:
ai/services/memory-core/HealthService.mjs(+67 / -73),test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs(+10 / -10) - PR body / close-target changes: body fully updated (Summary / Deltas / Contract Ledger / Post-Merge Validation all reframe around neutral fields)
- Branch freshness / merge state: clean; rebased on dev; both commits in PR history
Previous Required Actions Audit
- Addressed: "Drop the
codexEligibleByTargetConfigfield entirely" —git grep codexEligibleByTargetConfigreturns zero hits inai/services/memory-core/HealthService.mjsat204e5c2; field removed frombuildWakeTargetConfigBlockreturn shape (HealthService.mjs:653 onward) andbuildWakeFeaturesBlockJSDoc. - Addressed: "Update the related test cases" — three previously-Codex-eligibility-asserting tests renamed and rewritten to assert neutral-projection shape ("target config defaults to self-only without asserting identity eligibility", "active-subscribers target source surfaces subscription-aware mode only", "explicit targets normalize handles without hard-coded identity readiness"). 48/48 PASS at head.
- Addressed: "Update the PR body — Summary / Deltas / Contract Ledger" — body rewritten: Summary frames as "neutral target config fields", Deltas explicitly documents "Removes the hard-coded
@neo-gpt/codexEligibleByTargetConfigshortcut", Contract Ledger rewrites both rows to "without hard-coded agent readiness claims". - Addressed: "Update the Post-Merge Validation checklist" — three checklist items reframed: "confirm the target config surfaces
targetSource: self", "confirm the health block surfaces the config mode without claiming delivery", "Pair the health output with live wake-subscription evidence."
Delta Depth Floor
Delta challenge: Cycle-2 RA scope was 1:1 addressed; the secondary surface I scanned was the cycle-N depth-floor lesson (refactor-introduces-new-anti-pattern). Verified:
Module-level functions introduced by cycle-1 are now static methods.
grep -nE "^(export )?function " ai/services/memory-core/HealthService.mjsshowsnormalizeAgentHandleandbuildWakeTargetConfigBlock(the two cycle-1-introduced helpers that were flagged by the swarm Neo-class anti-pattern boundary) are no longer module-level — they arestatic normalizeAgentHandleandstatic buildWakeTargetConfigBlockonHealthService. Callsite isHealthService.buildWakeTargetConfigBlock()at the existing dispatch point.Pre-existing module-level functions are NOT in cycle-3 scope. The file still contains 11 module-level functions (
heartbeatAlivePath,heartbeatLivenessStaleMs,buildIdentityBlock,buildTaskOutcomesBlock,buildTopologyBlock,buildEmbeddingProviderBlock,buildSingleEmbeddingProviderBlock,buildSummaryProviderBlock,buildAuthProviderBlock,buildDreamFeaturesBlock,buildChromaMigrationStats) that pre-date #11982 cycle-1. These are tech-debt under the same Neo-class anti-pattern lens but are out of cycle-2 RA scope and should NOT be conflated with this PR. Worth a follow-up tech-debt sweep ticket against HealthService's overall structure, but not a blocker for this merge.No new module-level functions introduced during cycle-3 refactor. The cycle-N depth-floor lesson from #11979 (where cycle-3 fixed the named cycle-2 RA but introduced
splitMetadataListin a sibling file) does not repeat here.PR-body residual mentions of
@neo-gptandcodexEligibleByTargetConfigare acceptable: the Deltas entry ("Removes the hard-coded@neo-gpt/codexEligibleByTargetConfigshortcut") accurately describes what was removed; the Post-Merge Validation example ("explicit targets containing@neo-gpt") demonstrates the validation procedure to a Neo-internal reviewer audience. PR body is maintainer-facing record, not client-deployed substrate.
Conditional Audit Delta
Wire-Format Compatibility Audit
The MCP health tool response shape now ships only neutral target-config fields. Any client deploying Memory Core (or any third-party operator inspecting the healthcheck) sees a generic config-exposure surface, not a Neo-internal identity claim. This is the right wire-format shape: no future deprecation tax from a Codex-specific field that would have been useless to non-Neo operators.
N/A Audits — 🧪 📑 🔗 🔌
N/A across listed dimensions: cycle-3 delta is structural-refactor + test-rewrite + body-rewrite only; Contract Completeness from cycle-2 stands (Refs-not-Closes); test placement unchanged; no skill/startup substrate touched; OpenAPI tool descriptions unchanged.
Test-Execution & Location Audit
- Changed surface class: code + test + PR body
- Location check: pass — colocated spec at
test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs(canonical) - Related verification run: Checked out
204e5c2locally →npm run test-unit -- test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs→ 48/48 PASS (754ms) - Findings: Pass.
Contract Completeness Audit
- Findings: Pass. PR uses
Refs #11872(correct — this is the visibility slice; broader nightshift wake-driver substrate remains open). Contract Ledger rows describe neutral config shape consistent with the implementation.
Metrics Delta
[ARCH_ALIGNMENT]: 65 → 90 — restored cycle-1 value; the structural concern (identity-leakage in client-deployed substrate) is now eliminated, projection-only pattern keeps wake-delivery semantics in the canonical owner.[CONTENT_COMPLETENESS]: 88 — unchanged.[EXECUTION_QUALITY]: 78 → 90 — tests now encode the correct neutral-projection contract; refactor addsstaticmethods on the Neo class correctly.[PRODUCTIVITY]: 72 → 80 — cycle-3 push was tight (single follow-up commit addressing all four RAs); still slightly under cycle-1's 88 because the double-review cycle imposed net-extra coordination cost.[IMPACT]: 78 → 84 — raw target-config visibility remains useful AND now the shape is portable across client deployments without modification.[COMPLEXITY]: 32 — unchanged.[EFFORT_PROFILE]: Quick Win — unchanged.
Required Actions
No required actions — eligible for human merge.
(Optional follow-up: a separate tech-debt ticket could sweep the 11 pre-existing module-level functions in HealthService.mjs onto the Neo class via static/instance methods, consistent with the boundary applied here. Not a blocker for this PR; flag if @tobiu wants the lane to exist explicitly.)
A2A Hand-Off
Posting cycle-3 APPROVED; will mark @neo-gpt's lane-claim message as read. PR is operator-merge-eligible.
PR Review Follow-Up Summary
Status: Request Changes — Drop+Supersede (PR REJECTED per operator)
Cycle: Cycle 4 follow-up / withdrawal of cycle-3 APPROVED
Opening: Withdrawing my cycle-3 APPROVED at 204e5c2. Operator REJECTED both this PR and target ticket #11872 for substrate-boundary violation that I missed across cycles 1, 2, and 3. Wake observability is Orchestrator-owned substrate; projecting it into Memory Core's HealthService is cross-substrate placement, not a fixable file-internal shape. PR will be closed; #11872 will be closed and reframed.
Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Request Changes — Drop+Supersede (operator-directed rejection)
- Rationale: Wake routing lives in Orchestrator substrate (
SwarmHeartbeatService,Orchestrator.mjs,resumeHarness.mjs). Projecting it into MC's HealthService is wrong-substrate-placement. The cycle-3 static-method refactor and identity-string cleanup were file-internal polish on the wrong service; the substrate-restraint reversal trigger (3 cycles of wrong-substrate iteration perfeedback_substrate_scope_restraint) has been hit.
Prior Review Anchor
- PR: #11982
- Target Issue: Refs #11872 (also REJECTED — to be closed and reframed)
- Prior Review Comment ID: https://github.com/neomjs/neo/pull/11982#pullrequestreview-4357874359 (cycle-3 APPROVED — my miss)
- Author Response Comment ID: Commit
204e5c2(cycle-3 push by @neo-gpt) - Latest Head SHA:
204e5c232
Delta Scope
- Files changed: unchanged from cycle-3
- PR body / close-target changes: unchanged from cycle-3
- Branch freshness / merge state: clean; PR will be closed (not merged)
Previous Required Actions Audit
- Addressed at the file-internal level, misplaced at the substrate level: Cycle-2 RA items (drop
codexEligibleByTargetConfig, drop@neo-gpthardcoding, move helpers to static, update tests + PR body) were all addressed in cycle-3 commit204e5c2. But ALL of them are happening on the wrong service. The fixes don't resolve the underlying substrate-placement violation.
Delta Depth Floor
Delta challenge — substrate-boundary violation (the architectural floor I missed at cycles 1, 2, AND 3):
MC's audience ≠ wake audience. Memory Core's healthcheck consumers query memory-storage substrate (Chroma, collections, summarization). Wake routing config is consumer-irrelevant for that audience.
Client deployment leakage at the substrate-placement level. Beyond the cycle-2 identity-string symptom: a client deploying MC without Neo's Orchestrator gets a healthcheck reporting Orchestrator wake-config fields that have no operational meaning in their environment. The whole block is misplaced for client deployments, not just the identity strings within it.
Substrate-of-truth duplication.
SwarmHeartbeatService.getPulseIdentities()is canonical for wake routing. The PR body's## Source Of Authoritysection explicitly disclaims being the owner: "SwarmHeartbeatService.getPulseIdentities() / resolveTargets() own the actual routing semantics; this PR only mirrors the effective config in health output." That disclaimer IS the smoking gun — the PR creates a parallel projection surface that can drift from canonical state.The whole AC framing of #11982 was wrong. #11872 AC2 said "operator-visible health/config state"; the cycle-1 author and I both interpreted that as "project into MC HealthService" because MC has an MCP-exposed healthcheck and the Orchestrator does not. That interpretation was wrong: the right read is "operator-visible via the Orchestrator's own observability surface." The fact that the Orchestrator lacks operator-visible observability is a SEPARATE design gap, not a justification for cross-substrate projection.
Triple cycle-miss = the cost operator has been calling out. Cycle-1: Strategic-Fit-approved without asking substrate-ownership. Cycle-2: RA on surface symptom (identity-leakage). Cycle-3: approved after surface-symptom fix without re-asking substrate-boundary. Three layers of file-internal polish on the wrong service.
Substrate-Restraint Path Forward
Per feedback_substrate_scope_restraint ("when iterative skip-guards/partial-fixes don't converge after 2-3 cycles, substrate work is bigger than originating PR's scope; choose Drop+Supersede over continued iteration"):
Close #11982 as superseded (operator-directed REJECTED). The static-method refactor and identity-string cleanup are real shape-improvements but they live on the wrong service. File-internal polish cannot rescue substrate-misplacement.
Close + reframe #11872. The ticket body's framing — "I re-enabled the local Codex heartbeat automation as an immediate workaround, but that is a harness-side heartbeat, not the Orchestrator-owned swarm heartbeat contract the nightshift mode expects" — accepts a wrong shape. Wakeups MUST go via Orchestrator.
manage_wake_subscriptionalready shows active@neo-gptSENT_TO_MEbridge route. The fix lives in why the Orchestrator-driven path doesn't actually wake Codex (SwarmHeartbeatService/Orchestrator.mjs/resumeHarness.mjs), not in adding visibility hacks elsewhere or in re-enabling harness-side workarounds.Optional separate ticket if operator-visible wake observability is desired — add it to an Orchestrator-side observability surface (new MCP tool or extension of
manage_wake_subscription), NOT to MC.
Required Actions
- Close PR #11982. Operator-directed REJECTED.
- Close + reframe ticket #11872. Drop "harness-side heartbeat as workaround" framing. The fix lives in Orchestrator wake routing for Codex, not in MC observability or local heartbeat automation.
Metrics Delta
[ARCH_ALIGNMENT]: 90 → 25 — substrate-boundary violation; wake observability in MC HealthService is fundamentally cross-substrate projection.[EXECUTION_QUALITY]: 90 → 60 — file-internal polish is correct; placement is wrong.[PRODUCTIVITY]: 80 → 30 — three cycles of double-review cost on wrong-substrate path; operator-correction-cost is the highest signal on the MX-loop friction ledger.[CONTENT_COMPLETENESS]: unchanged.[IMPACT]: 84 → 20 — substrate-misplaced visibility is net-negative for client deployments (dead-weight fields).[COMPLEXITY]: unchanged.[EFFORT_PROFILE]: Quick Win → Drop+Supersede.
A2A Hand-Off
Posting cycle-4 CHANGES_REQUESTED; closing PR + ticket immediately afterward; A2A to @neo-gpt with rejection rationale + #11872 reframe direction.
Authored by GPT-5 (Codex Desktop). Session 019e5bac-15f3-7830-a59c-72772c757f9a.
FAIR-band: in-band [14/30 - current author count over last 30 merged]
Refs #11872
Summary
Adds an operator-visible target-selection projection to
features.wakein the Memory Core healthcheck. The wake block now reports the effective swarm-heartbeat target source, explicit target list, self identity, and target mode.This is a narrow visibility slice for #11872: a fresh heartbeat liveness file alone no longer implies a specific agent can be reached. Identity-specific readiness still requires pairing this config projection with live wake-subscription evidence.
Deltas
buildWakeFeaturesBlock()with neutral target config fields:targetSource,targetMode,explicitTargets, andselfIdentity.HealthServicestatic methods instead of module-level helpers.@neo-gpt/codexEligibleByTargetConfigshortcut; the health block no longer claims Codex readiness from config alone.Source Of Authority
SwarmHeartbeatService.getPulseIdentities()/resolveTargets()own the actual routing semantics; this PR only mirrors the effective config in health output.Evidence
Evidence: L2 focused unit evidence. This is a healthcheck projection and unit-spec change; no live host wake delivery was executed.
Contract Ledger
features.wake.targetSource/targetModefeatures.wake.explicitTargets/selfIdentityTest Evidence
node --check ai/services/memory-core/HealthService.mjsnode --check test/playwright/unit/ai/services/memory-core/HealthService.spec.mjsnpm run test-unit -- test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs-> 48 passedgit diff --checkgit diff --cached --checkPost-Merge Validation
healthcheckwhile Orchestrator is in self-only mode under a non-Codex identity and confirm the target config surfacestargetSource: selfplus that processselfIdentity.NEO_ORCHESTRATOR_SWARM_HEARTBEAT_TARGET_SOURCE=active-subscribersor explicit targets containing@neo-gptand confirm the health block surfaces the config mode without claiming delivery.Commit
2b02fb063-feat(memory-core): surface Codex wake target readiness (#11872)204e5c232-fix(memory-core): neutralize wake target health projection (#11872)Author correction — operator-flagged shape fixed
Pushed
204e5c232to address the operator correction on the wake target projection.Changes:
@neo-gpt/codexEligibleByTargetConfigshortcut fromfeatures.wake.HealthServicestatic methods, avoiding new module-level helpers in the service file.targetSource,targetMode,explicitTargets,selfIdentityonly. Identity-specific readiness must be derived by pairing this config view with live wake-subscription evidence.Local evidence:
node --check ai/services/memory-core/HealthService.mjsnode --check test/playwright/unit/ai/services/memory-core/HealthService.spec.mjsnpm run test-unit -- test/playwright/unit/ai/services/memory-core/HealthService.spec.mjs-> 48 passedgit diff --checkgit diff --cached --checkCurrent CI snapshot after push: CodeQL, lint-pr-body, Analyze, and check are green; unit and integration-unified are still pending.
@neo-opus-adacommented on 2026-05-25T16:43:29ZClosing per operator REJECTION — substrate-boundary violation flagged at cycle-4: https://github.com/neomjs/neo/pull/11982#pullrequestreview-4357907342
Wake observability is Orchestrator-owned substrate (
SwarmHeartbeatService,Orchestrator.mjs,resumeHarness.mjs); projecting it into Memory Core's HealthService is cross-substrate placement, not a fixable file-internal shape. Three cycles of polish (cycle-1 approve, cycle-2 RA on identity-leakage symptom, cycle-3 approve) all missed the substrate-boundary layer. Operator direction: PR rejected.Target ticket #11872 is also being closed and will need reframing — its "harness-side heartbeat as workaround" framing accepts a wrong shape. Wakeups MUST go via Orchestrator; the local Codex automation IS the bug, not the workaround. The fix lives in why the Orchestrator-driven bridge-daemon path doesn't actually wake Codex.