docs: scope "the runtime image" to ObjectOS where the claim depends on it

[os-bill]

Fixes #253

Scopes the bare phrase "the runtime image" to the ObjectOS runtime image at the four sites where the phrase is bare and the claim depends on which image is meant. Subject only — no behavioural claim is added, softened or removed anywhere in this diff.

The rule applied

Scope the phrase where the sentence's truth or the reader's action depends on which image is meant; leave it where the sentence holds for both. A bare phrase in a referent-independent sentence is not a defect, and renaming it would be churn.

What changed (4 sites, 4 files)

Site	Before	After	Why it is referent-dependent
deploy/air-gapped.mdx:84	"The runtime image lives in a private registry ..."	"The ObjectOS runtime image lives in ..."	The framework image is Apache-2.0 and public, so the sentence is simply false of it
reference/environment-variables.mdx:159	"The runtime image refuses a boot that carries a non-default OS_ARTIFACT_PATH"	"The ObjectOS runtime image refuses a boot ..."	The exact sentence whose unscoped subject produced the false conclusion in #251
reference/runtime-capabilities.mdx:59	"Pin the runtime image to one that includes every package ..."	"Pin the ObjectOS runtime image to one ..."	An actionable instruction whose answer differs by image
build/marketplace.mdx:60	"Pre-bundled with the runtime image"	"Pre-bundled with the ObjectOS runtime image"	What a default catalog ships with differs by image

Two paragraphs are re-wrapped to their file's prose width (77 and 70 columns respectively); that is the only reason the diff shows 7 insertions rather than 4.

Carve-outs — respected, and here is the check

• reference/environment-variables.mdx:159 — only the subject is made explicit. The refusal claim is byte-identical apart from the inserted word: no assertion that the refusal exists, no softening, no deletion. That question is [finding] reference/environment-variables.mdx lists OS_ARTIFACT_PATH as retired and boot-refusing, but the open-source runtime image presets it and os start reads it #251's and stays open.
• reference/cli.mdx:54-56 — not touched at all. It already writes "on the ObjectOS / runtime image" (wrapped across the line break), so it needed no edit.

The fault line held: this PR fixes who a sentence is about, never whether that sentence is true.

The four bare uses deliberately left

Site	Why it stays bare
architecture.mdx:120	Release-axis framing — true of either image
configure/runtime.mdx:25	Same claim, same reason
reference/environment-variables.mdx:89	"The runtime image and the app it serves are independent release axes" — the same claim as the two above, so it fails the test for the same reason
resources/license.mdx:104	Referent B, already scoped in-sentence by "in the framework repository"

environment-variables.mdx:89 is the judgment call worth naming: it sits in an edited file, so leaving it is a decision rather than an oversight. Scoping it would have made the corpus newly inconsistent — the identical release-axis claim scoped on one page and unscoped on two others.

Sweep finding: the population is 13 sites, not 12

The card enumerated its sites. A single-line grep reproduces that list exactly, but a multiline sweep finds one more:

deploy/docker.mdx:13-14 — "registry credentials for the private runtime / image" — wraps across a line break, so no line-oriented search can see it. Referent A, already scoped by "private", so it needs no edit. The same wrapping is why cli.mdx:55 looks absent from a grep for "ObjectOS runtime image" while in fact being the one already-correct site.

Population: 13 occurrences over 10 files. Enumeration under-counted by exactly the site a line-oriented method is blind to.

Locale handling

runtime-capabilities.mdx and marketplace.mdx each have 7 locale siblings; air-gapped.mdx and environment-variables.mdx have zero. Per AGENTS.md, English is edited and the siblings are left to the translation pass — the freshness gate lists them under Stale translations, which is the non-blocking category by design. No sibling file is modified, and no translation: frontmatter is hand-written. zh-Hant is unaffected because zh-Hans is untouched.

Gates

Run at 38cdd9a. Exit codes captured by redirecting to a file first, never read through a pipe. Each gate's own verdict line:

Gate	Its own conclusion
gen-zh-hant.mjs --check	zh-Hant: 73 generated file(s) match the zh-Hans sources byte for byte
check-translations.mjs	translations gate passed — the 4 pages' siblings appear only under Stale translations (non-blocking)
check-translation-ownership.mjs	This PR touches 0 translation artifact(s) and 4 other file(s)
check-translation-output.mjs --self-test	29 rule case(s) ... every rule demonstrated able to fail
check-translation-output.mjs --files	translation output gate passed (139 pre-existing finding(s) reported) — none in these files
check-node-floor.mjs	Every declared floor clears what the dependency tree requires
turbo run type-check	1 successful — cache miss, so it ran on this content
turbo run build	1 successful — cache miss
check-locale-surface.mjs	every advertised URL has a source file and every source file is advertised; 397 docs entries, composition unchanged
turbo run test	1 successful, but FULL TURBO cached — its declared inputs exclude content/docs, so it is not a measurement of this change

The build logs 465 Failed to load dynamic font errors from OG-image generation (self-signed certificate in certificate chain). Pre-existing environmental noise from the sandbox proxy: the count is identical (465) in a build of the parent commit, and the build exits 0.

Rendered-output check: all four new strings appear in apps/docs/.next/server/app/**, and none of the four replaced strings appears anywhere in the build.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ChPQM8jamxLUfUAxwFpJ8S