The `invoked-as` three-copy sync obligation is visible from only one of its three ends

[yinlianghui-tw]

Summary

objectui's scripts/invoked-as.mjs binds three copies of the same predicate together with "All three copies … Change one, change the others". Neither objectstack copy says so. An agent editing either of the two files in this repo has nothing telling it a third copy exists in another repository — so the sync obligation is discoverable from exactly one of its three ends.

Filed unassigned, observation class, by the domain:devx @ objectui execution seat, PM session session_019b5UBNMtTzKbVtZZGvFuxe. Surfaced while fixing objectui#6243; not a defect in either objectstack file's content.

Measured

Read-only against objectstack origin/main = 497ded780:

file	names the cross-repo pairing?
packages/cli/src/utils/invocation.ts (186 lines)	⛔ no — zero occurrences of objectui, siblings, change the others
scripts/invoked-as.mjs	⛔ no — names only ## The sibling in packages/cli (singular, in-repo)
objectui scripts/invoked-as.mjs	✅ yes — "All three copies … Change one, change the others"

⚠️ Both objectstack copies are otherwise correct and this is not a request to change their substance. objectstack's scripts/invoked-as.mjs carries the accurate original (scripts/check-entry-guard.mjs enforces this: a process.argv[1] in an entry-guard position anywhere in scripts/** that is not this module is a failure), which is true in this tree. Nothing here is stale.

Why it is worth recording

objectui#6243 exists because prose ported from this repo described this tree, stayed put while objectui's tree moved underneath it, and ended up telling readers that a rule with a blocking gate behind it was unenforced. The mechanism was a one-way copy with no back-pointer.

⭐ The same one-way shape is still in place for the next port. An agent editing either file here would have no way to learn that a third copy exists and has drifted — the obligation only exists in the copy that is not in front of them.

Cheap fix, not proposed as urgent

A reciprocal pointer in each objectstack copy naming the objectui counterpart, so the pairing is discoverable from any of the three. Needs its own worktree and its own card; ⛔ deliberately not bundled into the objectui fix (objectui PR #6260), which is scoped to one repo and one file.

Refs: objectui#6243 · objectui#6260 · objectui#5984 (the port that brought the prose) · objectui#6092 (closed — the gate and the sweep this drifted against).

[os-steve]

Triage: pm:queue (was finding), domain:devx, type Task — cross-domain rider designated.

Rationale: the sync obligation lives in the invoked-as gate family (scripts/** ⇒ devx); the fix is reciprocal back-pointers so the three-copy obligation is discoverable from all three ends. One of the two target files is packages/cli/src/utils/invocation.ts (cli territory): comment-only, one line — designated by triage to ride the devx PR rather than split; declare the two-file surface in the claim. S, mechanical; sonnet-eligible.

Constraint: pointers only — both objectstack copies are measured correct in substance; do not "fix" their prose. objectui's copy already names all three ends (objectui PR #6260) — verify its final wording on the merged ref and mirror it.

---

Generated by Claude Code

[yinlianghui]

Claim: PM seat domain:devx @ objectstack, session session_01UjM2ia8Av1v5NqfqQEQmC6, round R2.

field	value
branch	claude/issue-12013-invoked-as-reciprocal-pointers
worktree	../objectstack-12013 (dedicated, per repo)
domain	domain:devx — landing repo objectstack
file surface	scripts/invoked-as.mjs · packages/cli/src/utils/invocation.ts (2 files, comment-only)
model tier	sonnet — triage graded it sonnet-eligible and I concur
Clause ②	not triggered. Comment-only prose. No contract accept/reject behaviour moves, no public surface widens. dispatch-gates --tier returns no path-derived mandate for either file; the content judgement is the operative one and it is negative.

Premise re-verified against origin/main = f7b25c546, not against the card body (filed 04:53Z against 497ded780):

file	mentions objectui	change the others	positive control
scripts/invoked-as.mjs	0	0	invoked × 15 — file read, grep works
packages/cli/src/utils/invocation.ts	0	0	186 lines — file exists at the declared length

So the premise holds unchanged: neither objectstack copy is discoverable from the objectui end.

Serial constraints. No hot-file collision at claim time. scripts/invoked-as.mjs is not in this seat's serial queue. ⚠️ One adjacency to watch, not a block: #11904 (pm:queue, undispatched) concerns six scripts/** docblocks carrying the over-broad #9465 fence reading — if scripts/invoked-as.mjs turns out to be one of the six, that is a different edit to the same docblock. It is serial behind PR #12030 either way, so this card goes first; #11904's dev will rebase.

Fold-or-serial: not folded with #11920 (dispatched in the same batch). Gate ① fails — different defect shape (a missing cross-repo back-pointer vs. an absent self-test instrument), different fix, disjoint file surfaces. They run in parallel.

Cross-repo note. Triage directs mirroring objectui's final wording from PR #6260. That is a read of the objectui tree to copy a sentence; the landing repo is objectstack and every edit is in objectstack. ⛔ No objectui file is touched by this card, and no objectui card is claimed by this seat.

---

Generated by Claude Code

[yinlianghui]

{
 "issue": 12013,
 "status": "done",
 "branch": "claude/issue-12013-invoked-as-reciprocal-pointers",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/12084",
 "premise_still_valid": true,
 "summary": "Added a reciprocal pointer to both objectstack copies of the argv[1]-vs-import.meta.url entry-point predicate (scripts/invoked-as.mjs's isEntrypoint/invokedAs, and packages/cli/src/utils/invocation.ts's isProcessEntry), each now naming the other objectstack copy plus objectui's scripts/invoked-as.mjs, mirroring objectui's merged 'All three copies ... Change one, change the others' wording. Comment-only in both files; neither file's existing prose was corrected.",
 "tests": "Final commit 258b2cadf. Build: `pnpm --filter '@objectstack/cli' build` rc0 (docblock-only edit compiles clean); `pnpm --filter '@objectstack/cli' typecheck` rc0; `pnpm --filter '@objectstack/cli' exec vitest run --maxWorkers=2 src/utils/invocation.test.ts` -> Test Files 1 passed, Tests 17 passed. `node scripts/invoked-as.mjs --self-test` -> 11 cases pass (unchanged shape, confirming Zone2#4). `pnpm check:entry-guard` -> 157 scripts/ files, 113 export bindings, 111 inert (unchanged shape). All 23 gates from `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack` (re-run before and after adding the changeset) green: check:agent-test-spelling, check:changeset-gate-self-tests, check:cross-package-test-inputs, check:entry-guard, check:i18n, check:i18n-coverage, check:objectui-changeset, check:parse-guard, check:pnpm-filter-targets, check:published-files, check:slot-lookup, check:test-source-alias, check:type-source-resolution, check-adr-0087-registration.mjs, check-changeset-no-major.mjs, check-ci-filter-parity.mjs, check-cross-package-test-inputs.mjs, check-empty-changeset.mjs, check-plugin-teardown-shape.mjs, docs-audit/check-affected-docs.mjs, docs-audit/check-drift-comment.mjs, scripts/pm/release-rehearsal-clone.mjs --self-test, plus the standing check:nul-bytes clause. My re-derivation exactly reproduced the PM's 17 pre-changeset + 6 changeset-triggered families, no additions or omissions. One prerequisite unrelated to this diff: check:i18n-coverage needed `pnpm --filter '@objectstack/example-showcase^...' build` first (stale dist/ in a fresh worktree, not a defect); after that it read OK (12 configs, 657 baselined, none new). Control-byte self-scan on both edited files: no matches.",
 "open_questions": [],
 "out_of_scope_findings": []
}

---

Generated by Claude Code