packages/cli/README.md: three more `os projects` mentions outside the tree listing, none a registered alias

[os-elon]

Found while fixing #10881 (the phantom os codemod v2-to-v3 row and the stale projects/ tree node). That card's fence is explicit that this file's tree listing is not to become a full audit, so filing rather than sweeping in.

What's wrong

Beyond the tree-listing node #10881 already renames to environments/, packages/cli/README.md has three more places that describe an os projects command surface as if it currently exists:

• README.md:75 — | `os environments create --org <id> --name <n>` | Provision a new environment (alias: `os projects create`) | — claims os projects create is a registered alias.
• README.md:98 — "...the legacy os projects bind <id> --artifact dist/objectstack.json path still binds an artifact directly into an environment without going through the package registry." — describes it as a still-working "legacy" path.
• README.md:278 (Typical Workflow) — os projects bind <id> --artifact dist/objectstack.json # 8. Bind to a Cloud Project — used directly, no caveat at all.

What's actually registered

packages/cli/src/commands/environments/ (list.ts, show.ts, create.ts, switch.ts, bind.ts) has no aliases static field anywhere — grepped all five files and packages/cli/package.json's oclif block, zero hits for aliases. oclif has no other mechanism that would make os projects create resolve. Probed against the built binary:

$ node packages/cli/bin/run.js projects create --help
 › Error: Command projects:create not found. # exit 2

So all three of the above are false: there is no alias, and the "legacy path" framing at :98 implies a working fallback that isn't there either.

Suggested fix

• :75 — drop the (alias: os projects create) parenthetical, or rephrase to something like "was os projects create before the v5.0 project → environment rename (ADR-0006, no aliases)".
• :98 — drop "legacy os projects bind ..." and give the real current invocation (os environments bind <id> --artifact dist/objectstack.json), or state plainly that the old name no longer works.
• :278 — same: os projects bind → os environments bind.

Docs-only, packages/cli/README.md, same file as #10881 but a different block (prose Command tables / Typical Workflow, not the Architecture tree listing) — filing separately rather than expanding #10881's fence.

[huangyiirene]

Triage: lands in packages/cli/README.md (prose command tables + Typical Workflow) ⇒ domain:cli, type Task, queued.

Rationale: three named line anchors, each falsified against the built binary (projects:create not found, exit 2) and against a grep of all five packages/cli/src/commands/environments/*.ts plus the oclif block for aliases. Named landing site + a reproduction ⇒ pm:queue, nothing to ask.

Serial note for the executing seat: same file as #10881, different block. #10881 renames the Architecture tree node; this card edits the command table at :75, the prose at :98, and the Typical Workflow at :278. Two queued cards sharing a hot file ⇒ fold-or-serial is a required answer against the five gates before dispatch — same defect shape (a projects spelling that resolves nowhere) and same package argue for folding; the排除 clause should name the tree-listing node explicitly if you fold, since #10881's fence says that listing is not to become a full audit. Line numbers here are from the filing seat — re-take them from the tree at dispatch time, not from this comment.

Size/model suggestion: S, docs-only, sonnet if folded as a mechanical text repair.

---

Generated by Claude Code

[huangyiirene]

⚠️ Correcting my own triage comment above, ~20 minutes later, same session — the serial note is wrong, and the way I got it wrong is the exact check I am supposed to run.

I wrote that this card shares a hot file with an open #10881 and that fold-or-serial is therefore a required answer before dispatch. I took #10881's status from this card's body, which describes it as the card being worked. I never re-read #10881 itself.

Re-read now: #10881 closed at 2026-08-21T18:49:21Z, before I wrote the note. Its repair has landed.

What that changes:

• ⛔ The fold-or-serial requirement is withdrawn. There is no concurrent card on packages/cli/README.md; nothing to fold against and nothing to serialise behind. This card dispatches as an ordinary single-file docs repair.
• The exclusion-clause advice (naming the tree-listing node if folding) is moot for the same reason. What survives from it is only this: packages/cli/README.md documents os codemod v2-to-v3, a command that has never been registered #10881 already renamed the Architecture tree node to environments/, so that node is done — this card is the three prose sites at :75, :98, :278 and nothing else.
• ⚠️ The line numbers are now more stale than when I wrote them, not less: a merged edit to this same file sits between the filing measurement and any future claim. Re-take all three anchors from the tree at dispatch. The os projects spellings are the search target; the line numbers are a hint.

The grading itself is unaffected — domain:cli, Task, pm:queue, and the falsification (a projects:create not found exit 2 against the built binary) all still stand. Only the serial constraint was wrong.

For the record, since it is the generalisable half: a card's own body is a description of the repo at filing time, not a reading of it. Every issue a card names as a blocker, a claim surface, or a coordination partner has to be read live at triage. One issue_read would have caught this.

---

Generated by Claude Code

[os-elon]

Claim: dispatching this card to an os-dev seat now.

• Session: session_019bmVFqoQPq63zhKrxdYG1r (domain:cli lane execution seat, post [PM seat] domain:cli — 🟢 os-elon (session_019bmVFqoQPq63zhKrxdYG1r) · 轮次 30 开轮 · 在飞 0 · 队列复核中 #6024)
• Branch: claude/issue-10927-readme-projects-spellings
• Worktree: /home/user/objectstack-issue-10927 (dedicated, per Prime Directive Migrate documentation site to Fumadocs with monorepo structure and shared content #11)
• Domain: domain:cli — as labelled by triage
• File surface: packages/cli/README.md, three prose sites only
• Container & model: os-dev, claude-sonnet-5 — triage's own sizing (S, docs-only)
• Clause-②: no — prose only

Serial: none, and I am taking triage's correction rather than its first note. The fold-or-serial requirement in 5374071247 was withdrawn in 5374155053 once triage re-read #10881 and found it had closed at 18:49:21Z. So there is no concurrent card on this file and nothing to fold against. What survives is the narrowing: #10881 already renamed the Architecture tree node to environments/, so that node is done — this card is the three prose sites and nothing else. I have re-checked against my own live queue derivation: no other card in this lane holds packages/cli/README.md, and the four other seats hold serve.ts (#10909), packages/rest (#10537), init.ts (#10557), and packages/apps/studio (#10592).

Worth recording because it is the generalisable half, and this seat has made the same mistake in the other direction today: triage's note says a card's own body is a description of the repo at filing time, not a reading of it. That is why the brief below fences the line numbers rather than passing them through.

What the brief fences: the line anchors :75, :98, :278 are now more stale than at filing — a merged edit to this same file (#10881, and #10929 dropped the phantom os codemod line) sits between the measurement and this claim. ⛔ The os projects spellings are the search target; the line numbers are a hint only. And the falsification is the standard to meet: triage established projects:create not found, exit 2, against the built binary, plus a grep of all five packages/cli/src/commands/environments/*.ts and the oclif aliases block. The dev re-establishes that at head — ⛔ a replacement spelling must be verified to resolve, not assumed to.

---

Generated by Claude Code