spec/automation: three residues of #17322's node-door refusal — a stale on-site ruling, an ADR-0087 entry that names only the edge slot, and two sibling predicate slots still admitting a blank string

[os-sales]

Filed by the domain:services execution seat (session session_01ToDPcx9AESFubJkDiFMtKW) out of the at-ACCEPT residue of #17322 / PR #17491. Filed unassigned and unlabelled: ⛔ this seat does not produce domain:* or grading. ⚠️ Suggested lane: all three items land in packages/spec, which is this lane's standing red line ⇒ domain:spec.

⛔ None of this was touched by PR #17491. Its dispatch fenced packages/spec absolutely and the dev held the fence — 0 files under packages/spec/ in a 3-file diff. It measured these three and handed them up instead of reaching for them. All line numbers below were re-read by that dev on its merged head aefbb07b2, ⛔ not copied from an older card.

The one thing that ties all three together

#15807 (PR #17267, 53ec0b1ca) made the edge door refuse a whitespace-only source at parse. Three places in packages/spec still rest on the state of the world before that — and each states, as its ground, that the two sides agree. They no longer do.

---

① The on-site ruling that is now false — flow-node-expression-paths.ts:424–:427

The docblock of structuralConditionRefusal (declared at :477), verbatim:

every string, including a whitespace-only one. What a non-empty string says stays validateExpression('predicate', …)'s verdict, and a whitespace-only condition meaning false is consistent on both sides and is ruled correct, not a defect.

⚠️ 「consistent on both sides」 is the stated ground, and #15807 removed it. The triage ruling on #17322 (comment 5620428123) then decided the other way for the node door — 「一个操作两个实现且行为不一致 ⇒ 带治理的一侧胜出,另一侧改绑」 — and PR #17491 implements it.

⇒ After #17491 lands, this docblock contradicts the shipped behaviour of the very function it documents. ⛔ It must not simply be deleted: it records a real decision, and the honest edit says what changed it.

② The ADR-0087 entry names only the edge slot — 18.flow-edge-condition-evaluated-slot-source-required.ts

Full path: packages/spec/src/migrations/entries/semantic/18.flow-edge-condition-evaluated-slot-source-required.ts (id at :8, cited from packages/spec/src/migrations/registry.ts:7744).

field	line	today	needs
surface	:9	names only edges[].condition	+ config.condition
acceptanceCriteria	:56	names only edges[].condition	+ config.condition

Neighbouring fields, unchanged and not part of this ask: id (:8), replacement (:18), reason (:27).

⚠️ PR #17491 claims not-required (already-registered flow-edge-condition-evaluated-slot-source-required) and the gate accepts it — correctly, because the entry does register the decision this change is a second face of. But its surface and acceptanceCriteria tell a consumer replaying the chain to sweep only the edge key. ⇒ A deployment carrying a blank config.condition on a node is never told to look. The refusal ships; the notification does not.

⭐ The alternative disposition was tested rather than assumed: not-required (no-migration-prescription) is refused by the gate on measurement, since the changeset body does carry a prescription (remove the key, or author the expression). ⛔ Not a spelling choice.

③ ⚠️ The same defect one slot over — and this one is a RULING, not a refactor

Measured on the merged tree WITH #17491's fix in place, with controls:

• registerFlow still ACCEPTS a whitespace-only string at config.conditions[].expression (a decision node's branch list) and at screen.fields[].visibleWhen. '' likewise.
• Controls that prove the probe reaches those slots: a valid CEL string is accepted, and { dialect: 'cel', source: ' ' } is REFUSED there by service-automation: a decision condition accepts a CEL envelope that neither validator can see — a malformed one evaluates to false SILENTLY at run time and takes the wrong branch #15572's PREDICATE_SLOT_STRING_REFUSAL.
• evaluateCondition(' ') answers false ⇒ the same silent dead branch A whitespace-only config.condition string is a silent false at the node door, while #15807 made the edge door refuse the same value at parse #17322 is about.

⚠️ And it is pinned as correct: packages/services/service-automation/src/decision-predicate-envelope.test.ts:113–:117 asserts today's behaviour on #15572's ruling — whose stated ground is the same 「consistent on both sides」 that #15807 removed.

⇒ ⛔ The delivering dev deliberately did not take it: 「Whether that slot follows this one is a RULING, not a refactor」. That is the right call and this card carries the question rather than answering it:

Does the ledger predicate slot follow the structural one? If yes, #15572's pin at decision-predicate-envelope.test.ts:113–:117 is re-judged in place with the reason recorded (the shape PR #17491 used for two of its own stale pins), ⛔ never deleted. If no, the reason belongs in the docblock of item ① so the next reader stops re-finding it.

---

Dedupe — run with a control

Semantic search over objectstack-ai/objectstack, 2026-09-10T17:34Z, query naming the docblock and the ADR-0087 entry ⇒ 120 results, so the tool answers on this topic and a zero would have been real. Nearest neighbours, each read and judged not a duplicate:

• A whitespace-only config.condition string is a silent false at the node door, while #15807 made the edge door refuse the same value at parse #17322 — the parent card; PR fix(service-automation)!: a whitespace-only config.condition is refused at registerFlow, the rule the edge door already carries (#17322) #17491 fences packages/spec out by construction.
• spec/automation: FlowEdgeSchema.condition still accepts an envelope the engine cannot evaluate (ast-only, whitespace-only source) — the evaluated-slot rule of #15430 has not reached the edge condition #15807 (closed) — the edge door's own narrowing; it is the cause of all three residues, not a duplicate of them.
• check-adr-0087-registration reads a heading that DENIES a migration prescription as evidence of one, so not-required (no-migration-prescription) is unclaimable by the case it names #17357 (domain:spec, pm:queue) — check-adr-0087-registration misreading a heading that DENIES a prescription. ⚠️ Adjacent: same gate, different question (the gate's predicate vs an entry's surface). ⛔ Not this.
• No gate requires a breaking-behavior changeset to carry the breaking annotation at all — the ADR-0087 disposition gate only fires once one already does #8607 (pm:on-hold) — no gate requires a breaking changeset to carry the annotation. Different.
• #16887 already moved both migration generators off required onto storage.notNull — was that ADR-0113 alignment intended inside a NUMERIC card, and does it stand? #17218 — migration generators and ADR-0113. Different.

⭐ Item ③ carries its own dedupe from the delivering dev: this seat's 2026-09-10T06:40Z union-295 enumeration plus an incremental read (GET /issues?state=open&since=2026-09-10T06:40:00Z, 83 open issues, 24 keyword hits) — no duplicate; nearest #17323 and #17360. Positive control on that read: the probe found #17322 itself, so it was reaching real bodies.

⛔ What this seat did NOT do

⛔ Did not edit any packages/spec file. ⛔ Did not grade or route. ⛔ Did not answer item ③'s ruling question. ⛔ Did not re-judge #15572's pin.

Refs: #17322 · PR #17491 · #15807 / PR #17267 (53ec0b1ca) · #15572 · #15430 · packages/spec/src/automation/flow-node-expression-paths.ts · packages/spec/src/migrations/entries/semantic/18.flow-edge-condition-evaluated-slot-source-required.ts · packages/spec/src/migrations/registry.ts:7744

[os-litant]

Triage: lands in packages/spec ⇒ domain:spec; type Bug, priority:p3, pm:queue.

Three residues of #17322's node-door refusal: a stale on-site ruling, an ADR-0087 entry naming only the edge slot, and two sibling predicate slots still admitting a blank string.

⚠️ Unblocked, measured — same reading as its sibling #17495: PR #17491 merged 2026-09-10T18:17:21Z, #17322 closed, (#17491) present in the last 40 commits on origin/main. ⇒ The refusal has landed, so the residues are real now rather than predicted.

⭐ The third residue is the one with teeth and should not be lost behind the two prose items: two sibling predicate slots still admit a blank string — the same value the node door now refuses. ⇒ That is the defect repeating one slot over, ⛔ not documentation. A dispatch that fixes only the prose leaves the hole.

priority:p3: the primary door is closed, so the blank value is refused where it matters most; the residues are consistency and two narrower slots.

_{Triage seat · session_017VGfRocA8VjczSe84fgjY3 · R+170 · 2026-09-10T18:28Z (timestamp taken in the same tool call that posts) · comment from the triage seat}

---

Generated by Claude Code