Skip to content

[skills-path 人工合并] docs(skill): dogfood §2 — a drifted console dist now refuses to boot (#7752) - #7837

Draft
huangyiirene wants to merge 1 commit into
mainfrom
claude/issue-7752-skill-note
Draft

[skills-path 人工合并] docs(skill): dogfood §2 — a drifted console dist now refuses to boot (#7752)#7837
huangyiirene wants to merge 1 commit into
mainfrom
claude/issue-7752-skill-note

Conversation

@huangyiirene

Copy link
Copy Markdown
Collaborator

Part of #7752 — runner-doc companion to #7783. Not Fixes: the issue closes with the guard, not with this note.

Why it is its own PR

.claude/skills/** and skills/** changes are ADR-level and merge through the human channel (maintainer ruling, 2026-08-11). Carrying this nine-line note inside #7783 would hold the entire guard — code, tests, changeset — behind that channel. Split so each moves at its own speed. Diff is one file, nine added lines, no code.

What it says, and why it is needed

#7783 makes os dev refuse to mount /_console when packages/console/dist/.objectui-sha provably differs from the repo's .objectui-sha pin, so a stale bundle is unreachable rather than silently authoritative.

That changes what a runner sees: /_console/ 404s, / stops redirecting to it, and the banner advertises no Console URL. Without this note it reads as a product bug. The skill's §2 already warns that the vendored console may be stale — what it could not say, until now, is that the 404 is the guard talking. So §2 gains:

  • what the refusal looks like, and that it is a precondition to fix, not a defect;
  • the fix — pnpm objectui:build (rebuild at the pinned SHA), with objectui:refresh named as the wrong turn, since it re-bumps the pin to the local ../objectui HEAD;
  • pnpm check:console-sha for the same comparison without booting;
  • OS_ALLOW_CONSOLE_DRIFT=1 for a deliberate stale boot — with the consequence stated, that every console observation then describes a commit the repo does not pin.

The RUNNER.md trap-table row saying the same thing from the checklist side stays in #7783 (docs/qa/** is not on the human-merge path).

Sequencing

This note describes behaviour that only exists once #7783 lands, so it reads as a promise until then. Landing it after #7783 keeps the skill honest; landing it before is harmless — the drift condition it describes is rare and the remediation it names is correct either way.


Generated by Claude Code

…7752)

Runner-facing counterpart to the guard in #7783. That PR makes `os dev`
decline to mount `/_console` when `packages/console/dist/.objectui-sha`
provably differs from the repo's pin, so the stale bundle is unreachable
instead of silently authoritative.

Without this note the new behaviour reads as a product bug: the skill's §2
already warns that the vendored console may be stale, but a runner who hits a
404 `/_console/` and an absent Console line in the banner has no way to know
that is the guard talking. So §2 now says what the 404 means and what fixes
it — `pnpm objectui:build` (rebuild at the *pinned* SHA; `objectui:refresh`
would re-bump the pin to the local ../objectui HEAD), `pnpm check:console-sha`
for the same comparison without booting, and `OS_ALLOW_CONSOLE_DRIFT=1` for
the deliberate stale boot, with the consequence spelled out.

Split out of #7783 because `.claude/skills/**` merges through the human
channel per the maintainer's 2026-08-11 ruling; keeping it there would hold
the guard behind that channel.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LkvcB2Ei8Mpaa87N2US5wY
@vercel

vercel Bot commented Aug 11, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment
Project Deployment Actions Updated (UTC)
objectstack Ignored Ignored Aug 11, 2026 7:39pm

Request Review

Copy link
Copy Markdown
Contributor

Review (skills seat): content verified — awaiting a human merge (#7548)

Terminal state per #7548: skills-root PR — stays draft, visibly hanging; the maintainer reviews and merges by hand. This seat's round reports carry it on the "awaiting a human merge" line until then.


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation size/xs skip-changeset PR has no user-facing published change; bypasses the changeset gate

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants