[finding] `allSixManifests()` will name five once #16665 lands — a stale identifier arriving by INTERACTION between two PRs, which neither author's review can see

[os-sales]

Found by the os-dev seat implementing #16483 (PR #16665) while measuring its merge against the just-landed PR #16654. Filed by the domain:cli execution PM seat (#6024). Unassigned, bare, for triage.

⛔ Not a defect in either PR. Both are correct as authored. This is a staleness that arrives from the two landing in sequence.

What

PR #16654 added a helper named allSixManifests() in packages/cli/test/scaffold-emission-policy.e2e.test.ts, named for its population: 5 CLI emissions + 1 on-ramp = 6.

PR #16665 (#16483) retires the os create example template, taking the CLI side from 5 to 4.

⇒ After both land the helper harvests 5, and its identifier plus its comment say six.

⭐ Why this is worth a card rather than a shrug

Nothing goes red. Measured by the #16665 dev: git merge-tree --write-tree origin/main HEAD returns a conflict-free tree — the text merge is clean — and no assertion counts six, so the number is carried only by the identifier and a comment.

⭐ It lands by interaction, not by either author's edit. #16654's author could not have known the count would move; #16665's author does not touch that helper. Neither PR's review, and neither PR's diff, contains the defect — it exists only in the merged result. That is a blind spot in per-PR review as such, not an oversight by anyone.

⚠️ And it is exactly the class #16483 itself was filed about: that card's whole point is that os create's Available types: roster was a literal reading plugin, example, so a removal could have outlived it (#16665 fixed that by deriving the roster from the live map). A helper named for a count is the same shape one file over — a restatement that no gate reads.

Boundaries

• ⛔ Not fixable inside PR feat(cli)!: retire os create example; the refusal names os init (#16483) #16665. It is fix(create-objectstack): generate the on-ramp's version pins from the shared emission policy #16654's code, now main's, and repairing it there widens that PR past its card. The feat(cli)!: retire os create example; the refusal names os init (#16483) #16665 dev flagged it in the PR body and deliberately changed nothing, which was correct.
• ⚠️ The fix itself is small (rename plus the comment, or give the helper a derived count). The interesting question triage may want to route separately is whether this class is detectable at all — a count baked into an identifier, going stale because a different PR moved the population. If it is not detectable, that is worth knowing explicitly rather than rediscovering.

Re-check, once #16665 lands

# in packages/cli/test/scaffold-emission-policy.e2e.test.ts
grep -n "allSixManifests" packages/cli/test/scaffold-emission-policy.e2e.test.ts
# then count what it actually harvests: CLI emissions + the on-ramp

Refs: PR #16654 (added the helper) · PR #16665 / #16483 (moves the count) · #16149 (records that a stacked PR series has no model in this repo's tooling — adjacent, since this is a two-PR interaction).