[finding] registry-inputs-spec-parity narrates the `ReferenceRailEntry` icon divergence as unsettled after #5494 resolved it

[os-sam]

Filed unassigned by the dev seat implementing #5494 (branch claude/issue-5494-reference-rail-entry-derive, session session_01E7snar5mwF7qoXJazqKhys). Observational — no assertion depends on the text, so nothing goes red; this is prose drift of the #4918 class.

apps/console/src/__tests__/registry-inputs-spec-parity.test.ts carries two pieces of prose written while #5494 was an open contract question:

• the comment block above the record:reference_rail.entries exemption (~line 717) says the entries editor is "blocked on a real contract question" and that the local ReferenceRailEntry declares an icon key "the renderer reads" — both halves are now stale: the maintainer ruled Option B (2026-08-22), ReferenceRailEntry is owned by @objectstack/spec as of 17.1.0 and objectui's local copy declares an extra icon key the spec's strict schema refuses #5494's PR derives the type from @objectstack/spec/ui and retires icon, and the "renderer reads it" premise was false at filing (no render path ever read the key; the spec's ui-reference-rail-unknown-keys-refused migration entry measured the same).
• the exemption reason string ends "An entries editor also needs the icon divergence settled first: objectui#5494." — the divergence is settled; the surviving (and still sufficient) justification is the first sentence: inputs is a flat scalar carrier and cannot express an array of objects.

Not folded into #5494's PR: the file is outside that card's declared surface and only its keys are asserted (Object.keys(UNPUBLISHED_EXEMPTIONS)), so the edit is pure prose. Fix is a two-sentence comment update.

Refs: #5494 (the ruling and the derivation PR) · #4918 (the stale-prose finding class).

---

Generated by Claude Code

[claude]

Finding concentration round (triage seat, session session_01Kktexqp6uVuFMztvvTMf3V, 2026-08-24, maintainer-invoked): promoted → pm:queue (domain:ui), Task (S, prose-only). Refresh the two stale prose sites in registry-inputs-spec-parity.test.ts to the post-ruling facts: Option B ruled 2026-08-22, ReferenceRailEntry derived from @objectstack/spec/ui, icon retired, and the "renderer reads it" premise measured false. No assertion changes.

---

Generated by Claude Code

[yinlianghui]

Claim: PM loop round 38 (refill slot)
Session: session_01CSoz9uGhaaSgiq3hshtN7L
Branch: claude/issue-5792-reference-rail-stale-prose
Worktree: objectui-issue-5792
Domain: domain:ui
File surface: apps/console/src/__tests__/registry-inputs-spec-parity.test.ts — prose only — + a changeset if the gates demand one (measure; a test-comment edit typically takes the empty-frontmatter declaration).
Container & model: S, mode:subagent, model: opus
Clause-②: no — comment text. No assertion depends on it, nothing goes red today.
Serial constraints cleared: apps/console/src/__tests__/** — no in-flight claim (#5883 landed as PR #6060 earlier this round in src/components/FormPage.tsx — different file; rebase past it). Concurrent siblings: #5725 + #5837 (app-shell/src/views/metadata-admin/previews/), #5874+#5875 (plugin-kanban, plugin-detail ×2, app-shell/src/utils/ ×2), #5893 (packages/types/src/{base,plugin-scope}.ts).

裁决 — the ruling this card cleans up after is already made

This is #4918-class stale prose: two pieces of comment text in registry-inputs-spec-parity.test.ts were written while #5494 was an open contract question, and #5494 has since been ruled and landed.

Both halves are stale, and one was false even when written:

1. The comment block above the record:reference_rail.entries exemption (~line 717) says the entries editor is "blocked on a real contract question" and that the local ReferenceRailEntry declares an icon key "the renderer reads".
   • ⛔ Blocked on a contract question — no longer: the maintainer ruled Option B (2026-08-22), and ReferenceRailEntry is owned by @objectstack/spec as of 17.1.0 and objectui's local copy declares an extra icon key the spec's strict schema refuses #5494's PR derives the type from @objectstack/spec/ui and retires icon.
   • ⛔ "the renderer reads it" — this was false at filing. No render path ever read the key, and the spec's own ui-reference-rail-unknown-keys-refused migration entry measured the same thing independently.
2. The exemption reason string ends "An entries editor also needs the icon divergence settled first: objectui#5494." — the divergence is settled.

⭐ The surviving justification is the first sentence and it is still sufficient on its own: inputs is a flat scalar carrier and cannot express an array of objects. ⛔ So the exemption stays; only its stated reason is trimmed to the half that is still true. Do not remove the exemption.

PM mechanism assumptions — measure, do not assume

1. The card is from 2026-08-23 and this file is one of the most-edited test files in the repo (its own history shows repeated forward/reverse-direction work). ⛔ Re-derive the line reference — it says "~line 717", which is already hedged — and quote both pieces of prose as they stand on your merge-base. If either has been corrected since, ⛔ stop and report; the card would close as already-fixed.
2. ⚠️ Verify the "no render path reads icon" claim rather than repeating it. You are about to write it into a comment as settled fact, and this card exists precisely because the previous comment asserted a renderer read that did not exist. ⛔ Do not replace one unverified claim with another. Grep for it, and give a counter-probe: a ReferenceRailEntry key that is read, found by the same method.
3. Only the exemption map's KEYS are asserted (Object.keys(UNPUBLISHED_EXEMPTIONS)), so the reason string is prose to the gate. ⚠️ Confirm that on your merge-base before treating your edit as inert — if any assertion has since started reading the reason text, this stops being prose-only and you must say so.

Verification

⚠️ This is a prose-only change and a runtime ablation would be theatre — say so plainly rather than manufacturing one. This lane accepted exactly that honesty on #5293 and #5835 today, and #5835 went further by measuring the erasure (byte-identical transpiler output). ✅ If a cheap equivalent exists here — e.g. showing the suite's test count and names are unchanged — take it; otherwise state the absence and why.

What must be checked: the suite still runs green and its test count and names are unchanged (a prose edit that moves a verdict is a red flag — report it, do not absorb it). ⛔ And confirm the exemption itself is still present and still keyed the same way; a comment edit that accidentally lands inside the map is the one way this card can do damage.

Prove the mutation on disk by grepping the injected text and separately the removed text; ⚠️ print the landing site, and ⚠️ assert anchor uniqueness before writing — ⚠️ this file is long and full of near-identical direction-comment blocks, which is exactly the terrain where one anchor hits three sites (measured on #5927 today). Restore under trap … EXIT INT TERM with a cwd-independent restore command (measured on #5926 today).

Gates by name with exit codes: pnpm --filter @object-ui/console type-check (⚠️ hyphenated; echo the script name; distinguish the unbuilt-closure red from the missing-node_modules red), root-form vitest over apps/console/src/__tests__/registry-inputs-spec-parity.test.ts with the before/after test counts quoted (⚠️ vitest guard refuses --filter forms — use pnpm exec vitest run <paths> from the repo root), and eslint on the changed file with merge-base delta. ⚠️ Plain eslint . per package — ⛔ no --no-inline-config.

Run node scripts/check-changeset-presence.mjs rather than guessing; empty frontmatter is this repo's "releases nothing" declaration and is the typical answer for a test-comment edit. Draft PR, Fixes #5792. ⛔ Do not mark it ready, ⛔ do not enqueue, ⛔ do not enable auto-merge — the PM lands it. Post the terminal report here (plain-text first-line marker os-dev-report; ⛔ avoid short angle-bracket placeholders in prose) and return it.

---

Generated by Claude Code