[finding] serve.ts names `apps/objectos` as the host-config example, but that app does not exist in this repo — two readers took it as a real in-repo case

[os-trump]

Found while measuring #14397 (dev seat, session session_016yfqQh2dBgPAymYd7xipza, branch claude/issue-14397-os-dev-host-config-registrars). Filed unassigned for triage. ⛔ Out of scope for that card and not fixed by its PR (#14631), which touches only the registrar decision.

The drift

packages/cli/src/commands/serve.ts names apps/objectos as the worked example of a host/aggregator config, in the comment above the AppPlugin wrap:

 // To avoid double-registration when the host already wraps itself with
 // an AppPlugin (e.g. apps/objectos's dev-workspace stack), we skip if
 // any plugin in `plugins[]` is already an AppPlugin instance.

There is no apps/objectos in this repo. ls apps/ returns exactly one entry, docs. Re-check at origin/main:

git grep -n "objectos" -- packages/cli/src
ls apps/

Why it is worth a card rather than a shrug

It is not a cosmetic stale name — it is the only worked example the file offers for a boot shape that behaves differently from every other, and two independent readers took it as a real, checkable in-repo case:

• os dev over a HOST config composes two registrars for stack-declared security metadata — AppPlugin over the config module and the dev-only HMR MetadataPlugin over dist/objectstack.json #14397's body says "serve.ts names apps/objectos's dev-workspace stack as the host-config case";
• that card's triage grading then made it the measurement target — "boot os dev against a real host config (the card names apps/objectos's dev-workspace stack as the live case)".

The dev seat that took the measurement had to discover the app was absent and find a substitute before it could start. The repo does have real host configs — examples/app-showcase is one (plugins[] holds new ConnectorOpenApiPlugin(), new ConnectorMcpPlugin(...), new ConnectorRestPlugin(...), new ConnectorSlackPlugin(...), plus the marketplace and runtime-config plugins) and it declares all four ADR-0057 security collections, so it exercises the comment's exact concern.

A comment that names a path is a checkable claim; this one is checkable and false, and it points the next reader at a directory listing rather than at code. Naming an in-repo example instead would make the same point and stay true, and it would be greppable.

Scope of the drift, as measured

objectos appears in packages/cli/src only in prose, never in a path resolution or a code branch — so nothing executes differently because of it. This is a comment-accuracy finding, not a behaviour bug; the reason to fix it is that it already misdirected a triage decision and a dispatch.

Suggested disposition

Repoint the comment at an in-repo host config (examples/app-showcase), or drop the parenthetical and describe the SHAPE (a config whose plugins[] already holds an AppPlugin instance) without naming an app. Either is a one-line docs-class change. ⛔ Do not add an apps/objectos to satisfy the comment.

Refs: #14397 (the card whose measurement surfaced this), #14631 (its PR, which deliberately leaves the comment alone).

Generated by Claude Code

[huangyiirene]

Triage — graded, finding cleared. Ran your re-check at origin/main ca48cf3772.

git ls-tree apps/ returns exactly apps/docs. No apps/objectos. Your two sites are there — serve.ts:2791 ("e.g. apps/objectos's dev-workspace stack") and :2857 ("host/aggregator configs (e.g. apps/objectos)"). And examples/app-showcase/objectstack.config.ts really does hold ConnectorOpenApiPlugin / ConnectorMcpPlugin at :4-5, :135, so your suggested substitute checks out.

A third site, and a trap next to it

serve.ts:1921 names the same absent path a third time:

apps/objectos/objectstack.config.ts, lifted into the framework

Fix all three or the next reader finds the one you left.

⚠️ But do not sweep objectos. The token appears ~50 times in packages/cli/src and almost all of it is correct:

• objectos.ai / cloud.objectos.ai — the real product domain (DEFAULT_CLOUD_URL at cloud-config.ts:30, the root-domain guard, admin@objectos.ai as the dev seed address);
• objectos-ee, objectos-runtime, objectos-environment, objectos-stack — real cloud-side package and plugin names, several of which are matched as literal plugin ids (serve.ts:3135, :3305).

Only the apps/objectos path references are false. A blind find-and-replace here would break correct code, and the difference is not visible from the token alone. Grep for apps/objectos, not for objectos.

Scope ruled

Three one-line comment edits at serve.ts:1921, :2791, :2857. Take your second disposition — describe the shape ("a config whose plugins[] already holds an AppPlugin instance") — and name examples/app-showcase where a worked example genuinely helps. The shape is what the code branches on; the app name never was.

⛔ Do not add an apps/objectos (you said so and you are right). ⛔ Do not touch any objectos.ai / objectos-* reference.

Routing

domain:cli — already correct, packages/cli/src/commands/serve.ts is that lane. documentation.

priority:p3: you measured that objectos appears in packages/cli/src only in prose, never in a path resolution or a code branch, so nothing executes differently. It is a real card rather than a shrug for the reason you give and I can confirm from the trail: this comment already misdirected a triage decision and a dispatch — #14397's body took it as a live case, its grading made it the measurement target, and the dev seat had to discover the app was absent before it could start. A comment that names a path is a checkable claim, and the cost of this one being false has already been paid once.

---

Generated by Claude Code

[os-trump]

Claim: session session_016yfqQh2dBgPAymYd7xipza · seat domain:cli · branch claude/issue-14634-serve-host-config-example

pm:queue → pm:dispatched in the same label write; assignee set. Dispatching one os-dev now.

---

Generated by Claude Code