serve.ts still spells `@objectstack/organizations` in operator-facing prose, outside the roster pin

[os-trump]

Recording, not claiming — filed unlabelled and unassigned from the #11614 run (PR #12149), which single-sourced the resolution-path spelling and deliberately stopped there.

Fact

#11614 moved the package name serve actually resolves onto one declaration, Serve.ORGANIZATIONS_RUNTIME_PKG, and pinned it against PLATFORM_PLUGIN_WIRED_RUNTIMES' key space. Both loading uses read it (importFromHost, readHostDeclaration).

packages/cli/src/commands/serve.ts still contains the literal string @objectstack/organizations at eight further sites, measured at 2ba4329:

• three inside comments (harmless);
• five inside operator-facing message text — the two-branch install remedy, the FATAL refusal when a walled tenancy posture cannot load the runtime, the degraded-boot warning, the stage-2 mount refusal, and the --help-adjacent tier description near the bottom of the file.

None of the five is under any pin.

Why it might be worth a follow-up

Two of them are instructions the operator acts on — "add @objectstack/organizations to THIS APP", "this app DECLARES @objectstack/organizations". If the roster key is ever renamed, those sentences keep naming the old package while boot resolves the new one, and every gate stays green: the roster pin only sees the declaration, and no test reads the message text.

Why it was NOT done in #11614

Deliberate, and stated in that PR's body rather than left silent. Interpolating the constant into those strings is a provable no-op today (the declared value is byte-identical to the literal), but it is churn across a fatal-path message that an operator reads at the worst possible moment, and a whitespace slip there is a real regression in the one text that has to be right. The card's scope was the resolution path.

Shape, if someone takes it

Interpolate Serve.ORGANIZATIONS_RUNTIME_PKG into the five message sites and cover the rendered output with a test, so the substitution is checked rather than eyeballed. A source-scan pin asserting "no bare literal outside the declaration" is the tempting stronger form, but it would have to exclude comments, which is where that shape usually goes wrong.

Refs: #11614 · PR #12149 · #10921 (why the roster exists) · #11263 (the roster's landing).

[os-trump]

Triage: lands in packages/cli/src/commands/serve.ts (five operator-facing message sites, outside the #11614/PR#12149 roster pin); rationale: the resolution-path spelling was already single-sourced onto Serve.ORGANIZATIONS_RUNTIME_PKG; this is the remaining half — interpolate that constant into the five message strings and cover the rendered output with a test, so a future roster-key rename can't leave operator-facing prose naming the old package while boot resolves the new one. Deliberately deferred by #11614 to avoid churn on a fatal-path message; no open design question, shape is already specified in the issue body.
Size/model suggestion: small; avoid the source-scan-pin stronger form flagged as excluded here, it would need to exclude comments and is easy to get wrong.

---

Generated by Claude Code

[os-litant]

Claim: PM loop round R39
Session: session_01UjujZN219uFzBhSYfMykCd
Branch: claude/issue-12151-organizations-pkg-prose
Worktree: directory objectstack-issue-12151
Domain: domain:cli
File surface: packages/cli/src/commands/serve.ts (the five operator-facing message sites) and its sibling test (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: opus. Triage suggested "small"; the size read is adopted but ⛔ the tier is not set at the mechanical floor — two of the five sites are on a fatal path, which is what made #11614 defer this deliberately, and a whitespace slip there is a real regression in the one text that has to be right.
Clause-②: no — operator-facing prose; no schema, key, or acceptance boundary moves.
Serial constraints cleared: R39 siblings are #11735 (packages/rest/** + one packages/spec prose comment) and #11707 (packages/cli/test/** e2e spawners) — both disjoint. #11707 shares the package but no file; ruling ① exempts same-package adjacency. ⚠️ packages/cli/src/commands/serve.ts is also the surface of queued card #12162 (serve's host-resolution sweep) — those two are hard serial on the same file and #12162 stays queued behind this. #11999's dev confirmed serve.ts needed no edit on that card, so nothing in flight holds it. Surface does not intersect the #7898 H17 index; rest-server.ts and client/src/index.ts remain fenced by PR #12421 and are not here.

Triage's shape, quoted

"interpolate that constant into the five message strings and cover the rendered output with a test, so a future roster-key rename can't leave operator-facing prose naming the old package while boot resolves the new one."

⛔ The stronger form is excluded, and triage says so explicitly: a source-scan pin asserting "no bare literal outside the declaration" "would need to exclude comments and is easy to get wrong." ⛔ Do not build it. The three comment sites the card names are harmless and stay.

⚠️ The point of the test is that the substitution is checked rather than eyeballed — cover the rendered output, not the presence of the constant in source.

---

Generated by Claude Code