[finding] The blank template ships no app, so the Console renders nothing the scaffolder wrote — the README promises the opposite

[os-elon]

Found during the new-developer first-run journey audit, #10264. Ranked there as the single friction most likely to lose a stranger.

What a newcomer sees

Run the two commands the README puts front and centre, log into the Console, and the app you were just told about is not there.

• /_console/home → "Your apps 1", and that one app is Setup, a platform app.
• my_app_note — the object the scaffolder just wrote — appears nowhere in the navigation.
• It is live the whole time: GET /api/v1/data/my_app_note → 200, POST → 201.

After authoring a single App.create({...}) by hand, "Your apps" became 2 and the objects became reachable. So nothing is broken — the object simply has no route into the UI.

Why it lands where it does

packages/create-objectstack/src/templates/blank/src/ contains only objects/index.ts and objects/note.object.ts. There is no *.app.ts and no *.view.ts. Verified on origin/main:

packages/create-objectstack/src/templates/blank/src/objects/index.ts
packages/create-objectstack/src/templates/blank/src/objects/note.object.ts

An object without an App is served over REST but absent from Console navigation, which makes README.md:69 — "The Console renders the real app — records, boards, dashboards" — false for the app the scaffolder actually produces.

Why a newcomer cannot get themselves out of it

Nothing they can reach tells them an app is the missing piece. The repo README does not say it. The generated README.md does not say it. The generated AGENTS.md lists src/apps/ in its project-structure section but never says it is required for anything to render. The rule is stated only in skills/objectstack-ui/SKILL.md (~L581), which a human following the README has no reason to open.

The most likely reading of the screen is "it didn't work", and the most likely next action is closing the tab — with the skills bundle, the gates and the enforcement, all of which the audit found genuinely excellent, sitting unseen on the other side.

Adjacent gap worth noting for whoever takes this

os validate reported 1 Apps only once an app existed. It never warned that zero apps means zero navigable UI. Whatever shape the fix takes, that silence is the same gap seen from the gate side.

Filed by the PM loop from the audit in #10264. Not yet graded or routed.

[os-elon]

Claim: domain:cli lane execution seat
Session: session_019bmVFqoQPq63zhKrxdYG1r
Branch: claude/issue-10317-blank-template-ships-no-app
Worktree: objectstack-issue-10317
Domain: domain:cli
File surface: the blank template's own tree and the README text that describes it (⛔ re-derive the real paths at head — do not trust a path copied from the card). ⛔ Nothing under content/docs/releases/**.
Container & model: M, mode:subagent, model: opus
Clause-②: no — this changes what the scaffolder writes into a new project and what its README claims, not what any running server accepts, serves or refuses.
Serial constraints cleared: #10405 (rendered pnpm-workspace.yaml) and #10323 (the "Created files" listing) are dispatched in the same package right now. Same package is exempt per this seat's standing ruling; same file is hard serial — ⚠️ if your diff needs either of those files, STOP and report the collision rather than editing across the boundary.

⚠️ Read the fork before you pick a side

This card says the blank template ships no app, so the Console renders nothing, while the README promises the opposite. That is two defects wearing one title, and they have opposite fixes:

• (a) the template is right and the README lies ⇒ correct the README to describe an empty starting point honestly.
• (b) the README is right and the template is incomplete ⇒ ship a minimal app so the promise holds.

⛔ Do not choose by taste. Establish which one the template is for — read the template's own stated purpose, what init offers alongside it, and what the other templates do by contrast. A "blank" template that deliberately starts empty is a legitimate product choice, and (b) would then be adding unrequested scope to every project scaffolded from it.

Measure the symptom first: scaffold from the blank template with the real CLI, boot it, and look at the Console. Report what actually renders, not what the card predicts.

If the evidence genuinely does not settle (a) vs (b), that is a product decision — report the fork with what you measured and ⛔ open no PR for the half you cannot justify. premise_still_valid: false and a clean stop are both legitimate deliveries.

Fences

• ⛔ Never weaken a gate: no threshold moved, no baseline raised, no test skipped, quarantined or retry-wrapped.
• If your direction is (b), the app you add must be the minimum that makes the README true — ⛔ not a demo, not a showcase. Every line ships into every blank project forever.

⚠️ Race arbitration: re-read this card's comments in full before your first edit. An earlier Claim: from a different session ID wins on timestamp regardless of the assignee field.

---

Generated by Claude Code

[os-elon]

{
 "issue": 10317,
 "status": "done",
 "branch": "claude/issue-10317-blank-template-ships-no-app",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/10507",
 "premise_still_valid": false,
 "summary": "The symptom is real and reproduced on a real scaffold-and-boot; the card's stated PREMISE is not. Measured with the real CLI against published 17.1.0 packages: `GET /api/v1/meta/app` returns 2 apps (setup, account — both platform, card said 1, stale but same substance: zero from the project) while `GET /api/v1/data/my_app_note` serves the scaffolded object 200 all session. But `README.md:69` does not promise what the card reads it as promising — it is step 3 of a three-step loop whose step 2 has the agent author the list view AND a nav group, so 'the Console renders the real app' is about step 2's output, not the scaffolder's. The fork resolves to (a) on four independent readings: the registry describes blank as 'Minimal starter — one object, REST API, ready to extend'; all three `os init` templates write src/objects/ only, including the one described as 'Full application with objects'; blank is the SOLE template (five remote ones retired), so anything added ships into every project forever; and no *.app.ts has ever existed under src/templates/** in this repo's history. I therefore did NOT implement (b). What I shipped is the half both readings agree on and the card names in its own 'Adjacent gap' section: nothing the newcomer can reach says an app is the missing piece, while `pnpm dev` prints the Console URL on every boot. Documentation only — one 'The Console' section in the generated README naming the path, the consequence and src/apps/*.app.ts as the remedy; zero change to what the scaffolder writes into src/. Scoped `Part of`, not `Fixes`, so the (b) product call stays open for the maintainer.",
 "tests": "All at final commit 1776d515c, clean tree (dispatch-gates: 'committed 3, working tree 0, untracked 0'). (1) RUNTIME REVERSE-VERIFICATION from real scaffold-and-boot runs: `create-objectstack my-app -t blank` (real CLI, real npm install of 17.1.0), `objectstack dev --ui --seed-admin`. BEFORE: `GET /api/v1/meta/app` -> 200, 2 items [setup, account]; `GET /api/v1/data/my_app_note` -> 200 {\"records\":[],\"total\":0}. AFTER adding ONE *.app.ts to the SCAFFOLDED PROJECT (never to the template) and rebooting: 3 items [my_app_main, setup, account], the project app carrying nav_notes -> objectName: my_app_note. Direction as predicted: the empty src/apps/ is the whole cause, an app is the whole remedy — which is exactly what the shipped README section instructs. (2) TEST ABLATION: reverted ONLY the README to origin/main, kept the new test file, re-ran it — RED as predicted: '× discloses that no app means no Console navigation / AssertionError: README must name the Console path: expected ... to match /\\/_console\\// ' with 'Test Files 1 failed (1) · 4 tests | 1 failed'. The three tree/vacuity assertions stayed GREEN, so the failure is the disclosure and not a broken fixture. README then restored from the commit and confirmed byte-identical (`git diff --exit-code HEAD` clean). ABLATION REBUILD STATEMENT: no build participates in this ablation and none is needed — the test reads packages/create-objectstack/src/templates/blank/README.md and the template tree from SOURCE at run time; nothing resolves through any dependency's dist/*.d.ts, so there is no stale-artifact path that could hold it falsely green. (3) PACKAGE: `pnpm --filter create-objectstack typecheck && pnpm --filter create-objectstack test` -> exit 0, 'Test Files 8 passed (8) · Tests 88 passed (88)' (7 pre-existing files + mine). (4) GATES re-derived with `node scripts/pm/dispatch-gates.mjs` (NO path arguments) after the final commit — the brief named NO gate family, so all are newly derived; each quoted line is the gate's own verdict, exit codes captured before any pipe. Path-matched: check:changeset-gate-self-tests '✓ check-empty-changeset --self-test: 118 assertions over real temp git repos'; check:objectui-changeset '✓ objectui-range --self-test: all checks passed'; check:slot-lookup '✓ slot-lookup ratchet holds: 107 unswept site(s) in 25 file(s), none new'; check:template-version-sync '✓ sync-template-versions --self-test: 40 assertions over temp fixtures'; check-adr-0087-registration.mjs '✓ this PR adds no declared-breaking changeset (1 non-breaking changeset(s) seen).'; check-changeset-no-major.mjs '✓ This diff introduces no `major` bump.'; check-empty-changeset.mjs '✓ No empty-frontmatter changeset introduced by this diff (1 declaring changeset(s) added).'; docs-audit/check-affected-docs.mjs '✓ affected-docs self-test: 262 cases pass.'. Convention-triggered by adding a test file: check:query-options-erasure '✓ query-options-erasure ratchet holds: 67 unswept non-test site(s) in 17 file(s), none new'; check:engine-double-contract 'check-engine-double-contract: OK — 340 pinned, 133 in the DEBT ledger, 2 exempt.'; check:where-matcher '✓ where-matcher conformance holds: 266 matcher(s) discovered … 0 silently-wrong and 0 unjudged … none new.'; check:type-check-coverage 'check-type-check-coverage: OK — 64/77 workspace packages type-checked (plus the root), 13 in the DEBT ledger, 1 exempt.'. Plus check:nul-bytes 'check-nul-bytes: OK (scanned 6145 text file(s) … no raw ASCII control bytes).'. DECLARED NARROWING 1 — check:type-check-debt NOT MEASURED locally: its --re-measure refuses on this worktree ('55 workspace dependenc(ies) of the ledgered packages have no built type entry point on disk'), which means not-measured and never not-applicable. It needs a full 77-package turbo build and the shared lock had two agents queued behind me, disproportionate for a three-file docs change. Bounded: create-objectstack carries NO entry in DEBT or TEST_DEBT and --re-measure only re-runs tsc for ledger entries, so a new test file here cannot move a ledger number; the structural half my change COULD move ran green above, and the file is provably inside the package's tsc program (tsconfig include:['src'], excluding only src/templates; package typecheck passes with the file present). DECLARED NARROWING 2 — one ablation run was executed OUTSIDE the shared verify lock after two acquisition attempts (85s wait, then a 9m00s 'VERDICT queue-timeout (exit 99) · never acquired'); it is a single no-build vitest file reading two files from disk, run with --maxWorkers=1, not heavy verification. CI runs the full farm regardless.",
 "open_questions": [
 {
 "question": "Should the `blank` template ship a minimal app after all — direction (b) — or stay objects-only as measured? I resolved this to (a) on the evidence and shipped documentation accordingly, but the final call on template CONTENT is a product decision, not mine.",
 "options": [
 "A — Keep blank objects-only (shipped). Backed by four independent readings: registry description names REST not the Console; all three `os init` templates write src/objects/ only, including the one described as 'Full application with objects'; blank is the sole surviving template so any addition ships into every project forever; and no *.app.ts has ever existed under src/templates/** in repo history. Cost: a newcomer who skips the README's step 2 still meets an empty Console — now explained rather than unexplained.",
 "B — Ship a minimal *.app.ts (+ a list view) in the template. Makes the Console non-empty on first boot with zero authoring. Cost: contradicts the template's own stated purpose and its 'clean slate' README framing, makes the registry description false, pre-writes metadata into every project ever scaffolded that nobody deletes, and undercuts the product's central pitch that the AGENT composes the app in step 2."
 ],
 "recommendation": "A. On the four fixed axes — Real business need: no measured pull for a pre-written app; the pull found in the audit was for an EXPLANATION, and blank's actual consumers reach it via REST/MCP, which works today. Long-term soundness: (b) is the workaround shape — it papers over a documentation gap by mutating a product default, and it makes two other surfaces (registry description, README framing) false in the same act. AI-authored-metadata safety: a pre-written app is scaffolded code an agent will treat as an authored intent and extend rather than replace, which is exactly how unrequested structure propagates; an empty src/apps/ plus an explicit rule is the stricter, harder-to-get-wrong starting point. Startup scope discipline: (b) is capability expansion with no pull, shipping forever into every project — the 2026-08-04 focus principle says default-tight. The axes do not conflict here; all four point to A."
 },
 {
 "question": "Sequencing, for the PM — not a collision, but it needs an ordering decision. Sweep #10353 has two members that edit the SAME file this PR touches (packages/create-objectstack/src/templates/blank/README.md): #10321 (doc-hostname convergence) and #10322 (one package manager everywhere). Both are unassigned, no branch, no worktree — so nothing was in flight and there was nothing to serialize against — but whichever lands second must merge. Separately, #10353's exclusion list drops #10317 as a 'behaviour change to template content', and my measurement falsifies that premise: the surviving half of #10317 is TEXT, which is precisely #10353's declared defect class.",
 "options": [
 "A — Land this PR first, then let the #10353 sweep rebase over it. My new section uses `pnpm`, matching the file as it stands today; if #10322 converges on `npm`, the section moves with the rest of the file in that same sweep.",
 "B — Hold this PR and fold its README section into the #10353 sweep as a seventh member, since the measurement puts it in that sweep's defect class."
 ],
 "recommendation": "A. The sweep is queued and unassigned with no branch, so holding this PR trades a landed, independently-verified fix for an unbounded wait. The merge surface is one inserted section in a file region neither member restructures, and the #10353 exclusion rationale should be corrected either way — that correction is PM bookkeeping, not a reason to block."
 }
 ],
 "out_of_scope_findings": [
 "filed as #10504: `os validate` drops the whole `UI:` row at zero apps (measured: `UI: 1 Apps` appears with an app, and the row is ABSENT — not `0` — without one, both exiting 0), so 'no navigable UI' and 'this summary does not report on UI' read identically. Filed unassigned with `finding` + `domain:cli`, no `pm:queue`. Note it also records evidence AGAINST the shape #10317 suggested for it: a warning would fire on every clean scaffold, because zero apps is the intended blank state — printing `UI: 0 Apps` is the cheaper reading."
 ]
}

---

Generated by Claude Code

[os-elon]

os-dev-report

{
 "issue": 10317,
 "status": "done",
 "branch": "claude/issue-10317-blank-template-ships-no-app",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/10507",
 "premise_still_valid": false,
 "summary": "The symptom is real and reproduced on a real scaffold-and-boot; the card's stated PREMISE is not. Measured with the real CLI against published 17.1.0 packages: `GET /api/v1/meta/app` returns 2 apps (setup, account — both platform; card said 1, stale but same substance: zero from the project) while `GET /api/v1/data/my_app_note` serves the scaffolded object 200 all session. But `README.md:69` does not promise what the card reads it as promising — it is step 3 of a three-step loop whose step 2 has the agent author the list view AND a nav group, so 'the Console renders the real app' is about step 2's output, not the scaffolder's. The fork resolves to (a) on four independent readings: the registry describes blank as 'Minimal starter — one object, REST API, ready to extend'; all three `os init` templates write src/objects/ only, including the one described as 'Full application with objects'; blank is the SOLE template (five remote ones retired), so anything added ships into every project forever; and no *.app.ts has ever existed under src/templates/** in this repo's history. I therefore did NOT implement (b). What I shipped is the half both readings agree on and the card names in its own 'Adjacent gap' section: nothing the newcomer can reach says an app is the missing piece, while `pnpm dev` prints the Console URL on every boot. Documentation only — one 'The Console' section in the generated README naming the path, the consequence and src/apps/*.app.ts as the remedy; zero change to what the scaffolder writes into src/. Scoped `Part of`, not `Fixes`, so the (b) product call stays open for the maintainer.",
 "tests": "All at final commit 1776d515c, clean tree (dispatch-gates: 'committed 3, working tree 0, untracked 0'). (1) RUNTIME REVERSE-VERIFICATION from real scaffold-and-boot runs: `create-objectstack my-app -t blank` (real CLI, real install of published 17.1.0), `objectstack dev --ui --seed-admin`. BEFORE: `GET /api/v1/meta/app` -> 200, 2 items [setup, account]; `GET /api/v1/data/my_app_note` -> 200 with total 0. AFTER adding ONE *.app.ts to the SCAFFOLDED PROJECT (never to the template) and rebooting: 3 items [my_app_main, setup, account], the project app carrying nav_notes -> objectName my_app_note. Direction as predicted: the empty src/apps/ is the whole cause, an app is the whole remedy — exactly what the shipped README section instructs. (2) TEST ABLATION: reverted ONLY the README to origin/main, kept the new test file, re-ran it — RED as predicted: 'x discloses that no app means no Console navigation / AssertionError: README must name the Console path', 'Test Files 1 failed (1) · 4 tests | 1 failed'. The three tree/vacuity assertions stayed GREEN, so the failure is the disclosure and not a broken fixture. README then restored from the commit and confirmed byte-identical (`git diff --exit-code HEAD` clean). ABLATION REBUILD STATEMENT: no build participates and none is needed — the test reads the template README and tree from SOURCE at run time; nothing resolves through any dependency's dist/*.d.ts, so there is no stale-artifact path that could hold it falsely green. (3) PACKAGE: `pnpm --filter create-objectstack typecheck && pnpm --filter create-objectstack test` -> exit 0, 'Test Files 8 passed (8) · Tests 88 passed (88)' (7 pre-existing + mine). (4) GATES re-derived with `node scripts/pm/dispatch-gates.mjs` (NO path arguments) after the final commit — the brief named NO gate family, so all are newly derived; each quoted line is the gate's own verdict, exit codes captured before any pipe. Path-matched: check:changeset-gate-self-tests 'check-empty-changeset --self-test: 118 assertions over real temp git repos'; check:objectui-changeset 'objectui-range --self-test: all checks passed'; check:slot-lookup 'slot-lookup ratchet holds: 107 unswept site(s) in 25 file(s), none new'; check:template-version-sync 'sync-template-versions --self-test: 40 assertions over temp fixtures'; check-adr-0087-registration.mjs 'this PR adds no declared-breaking changeset (1 non-breaking changeset(s) seen).'; check-changeset-no-major.mjs 'This diff introduces no major bump.'; check-empty-changeset.mjs 'No empty-frontmatter changeset introduced by this diff (1 declaring changeset(s) added).'; docs-audit/check-affected-docs.mjs 'affected-docs self-test: 262 cases pass.'. Convention-triggered by adding a test file: check:query-options-erasure 'query-options-erasure ratchet holds: 67 unswept non-test site(s) in 17 file(s), none new'; check:engine-double-contract 'check-engine-double-contract: OK — 340 pinned, 133 in the DEBT ledger, 2 exempt.'; check:where-matcher 'where-matcher conformance holds: 266 matcher(s) discovered ... 0 silently-wrong and 0 unjudged ... none new.'; check:type-check-coverage 'check-type-check-coverage: OK — 64/77 workspace packages type-checked (plus the root), 13 in the DEBT ledger, 1 exempt.'. Plus check:nul-bytes 'check-nul-bytes: OK (scanned 6145 text file(s) ... no raw ASCII control bytes).'. DECLARED NARROWING 1 — check:type-check-debt NOT MEASURED locally: its --re-measure refuses on this worktree ('55 workspace dependenc(ies) of the ledgered packages have no built type entry point on disk'), which means not-measured and never not-applicable. It needs a full 77-package turbo build and the shared lock had two agents queued behind me, disproportionate for a three-file docs change. Bounded: create-objectstack carries NO entry in DEBT or TEST_DEBT and --re-measure only re-runs tsc for ledger entries, so a new test file here cannot move a ledger number; the structural half my change COULD move ran green above, and the file is provably inside the package's tsc program (tsconfig include ['src'], excluding only src/templates; package typecheck passes with the file present). DECLARED NARROWING 2 — one ablation run was executed OUTSIDE the shared verify lock after two acquisition attempts (85s wait, then a 9m00s 'VERDICT queue-timeout (exit 99) · never acquired'); it is a single no-build vitest file reading two files from disk, run with --maxWorkers=1, not heavy verification. CI runs the full farm regardless.",
 "open_questions": [
 {
 "question": "Should the `blank` template ship a minimal app after all — direction (b) — or stay objects-only as measured? I resolved this to (a) on the evidence and shipped documentation accordingly, but the final call on template CONTENT is a product decision, not mine.",
 "options": [
 "A — Keep blank objects-only (shipped). Backed by four independent readings: registry description names REST not the Console; all three `os init` templates write src/objects/ only, including the one described as 'Full application with objects'; blank is the sole surviving template so any addition ships into every project forever; and no *.app.ts has ever existed under src/templates/** in repo history. Cost: a newcomer who skips the README's step 2 still meets an empty Console — now explained rather than unexplained.",
 "B — Ship a minimal *.app.ts (+ a list view) in the template. Makes the Console non-empty on first boot with zero authoring. Cost: contradicts the template's own stated purpose and its 'clean slate' README framing, makes the registry description false, pre-writes metadata into every project ever scaffolded that nobody deletes, and undercuts the product's central pitch that the AGENT composes the app in step 2."
 ],
 "recommendation": "A. On the four fixed axes — Real business need: no measured pull for a pre-written app; the pull found in the audit was for an EXPLANATION, and blank's actual consumers reach it via REST/MCP, which works today. Long-term soundness for THIS project: (b) is the workaround shape — it papers over a documentation gap by mutating a product default, and it makes two other surfaces (registry description, README framing) false in the same act. Making AI-written metadata hard to get wrong: a pre-written app is scaffolded code an agent will read as authored intent and extend rather than replace, which is exactly how unrequested structure propagates; an empty src/apps/ plus an explicit stated rule is the stricter, harder-to-get-wrong starting point. Startup scope discipline: (b) is capability expansion with no pull, shipping forever into every project — the 2026-08-04 focus principle says default-tight. The axes do not conflict here; all four point to A."
 },
 {
 "question": "Sequencing, for the PM — not a collision, but it needs an ordering decision. Sweep #10353 has two members that edit the SAME file this PR touches (packages/create-objectstack/src/templates/blank/README.md): #10321 (doc-hostname convergence) and #10322 (one package manager everywhere). Both are unassigned, no branch, no worktree — so nothing was in flight and there was nothing to serialize against — but whichever lands second must merge. Separately, #10353's exclusion list drops #10317 as a 'behaviour change to template content', and my measurement falsifies that premise: the surviving half of #10317 is TEXT, which is precisely #10353's declared defect class.",
 "options": [
 "A — Land this PR first, then let the #10353 sweep rebase over it. My new section uses `pnpm`, matching the file as it stands today; if #10322 converges on `npm`, the section moves with the rest of the file in that same sweep.",
 "B — Hold this PR and fold its README section into the #10353 sweep as a seventh member, since the measurement puts it in that sweep's defect class."
 ],
 "recommendation": "A. The sweep is queued and unassigned with no branch, so holding this PR trades a landed, independently-verified fix for an unbounded wait. The merge surface is one inserted section in a file region neither member restructures, and the #10353 exclusion rationale should be corrected either way — that correction is PM bookkeeping, not a reason to block."
 }
 ],
 "out_of_scope_findings": [
 "filed as #10504: `os validate` drops the whole `UI:` row at zero apps (measured: 'UI: 1 Apps' appears with an app, and the row is ABSENT — not 0 — without one, both exiting 0), so 'no navigable UI' and 'this summary does not report on UI' read identically. Filed unassigned with `finding` + `domain:cli`, no `pm:queue`. It also records evidence AGAINST the shape #10317 suggested for it: a warning would fire on every clean scaffold, because zero apps is the intended blank state — printing 'UI: 0 Apps' is the cheaper reading."
 ]
}

Supersedes comment 5364071344 — that one was posted with the <!-- os-dev-report --> HTML-comment marker, which GitHub's sanitizer stripped on store, leaving it invisible to a marker scan. This comment carries the identical report behind the literal-text marker instead. No content differs.

---

Generated by Claude Code

[os-elon]

pm-dispatch-outcome

Dispatch outcome: premise_still_valid: false — the symptom is real, the card's stated cause is not. PR #10507, scoped to the half both readings agree on.

I verified the falsification, and it holds

The load-bearing claim is that README.md:69 does not promise what this card reads it as promising. I re-derived it rather than accepting it — and initially checked the wrong file, which is worth recording because it is the same trap the card fell into: I looked at packages/create-objectstack/src/templates/blank/README.md (the scaffolded project's README), found no three-step loop, and nearly reported the citation as wrong.

The citation is to the root README.md, and it is exact:

• README.md:57-58 (step 2): "Add an 'Open tickets' list view and a Support nav group. Run npm run validate when you're done." — addressed to the agent.
• README.md:69 (step 3): "The Console renders the real app — records, boards, dashboards."

So line 69 describes the output of step 2, which the agent authors. It is not a claim about what the scaffolder writes. The card read a three-step loop's payoff line as a standalone promise about init output. That is a subtle misread and catching it is the single most valuable thing in this delivery.

Separately confirmed on origin/main: the template README contains zero occurrences of "Console", describes itself as "Minimal ObjectStack environment — a clean slate for building", and its Getting-started section teaches REST and MCP — not the Console. The template never promised a populated Console either.

The fork resolved to (a), on four independent readings

I fenced this as two defects with opposite fixes and said ⛔ not to choose by taste. The dev resolved it with evidence, not preference:

1. the registry describes blank as "Minimal starter — one object, REST API, ready to extend";
2. all three os init templates write src/objects/ only — including the one described as "Full application with objects";
3. blank is now the sole template (five remote ones retired), so anything added ships into every project forever;
4. no *.app.ts has ever existed under src/templates/** in this repo's history.

⭐ Reading (4) is the one that settles it. "Has never existed in repo history" turns a design-intent argument into a measurement.

What shipped, and why this is Part of rather than Fixes

Documentation only — one "The Console" section in the blank template's README naming the path, the consequence, and src/apps/*.app.ts as the remedy, plus a bidirectional pin test. Zero change to what the scaffolder writes into src/. That is the half both readings agree on, and the card names it in its own "Adjacent gap" section.

⛔ Direction (b) — shipping a minimal app in the template — was not implemented, correctly: it is a product decision about template content and is the maintainer's, not this seat's and not the dev's. The PR is scoped Part of, so it will not close this card and (b) stays open. If the maintainer wants (a) ratified, this card can be closed by hand; the argument for A is in the report and the cost of B is stated there too (it would make the registry description false and pre-write metadata into every project forever).

Reverse-verified on a real scaffold-and-boot: before, GET /api/v1/meta/app returns 2 platform apps and nothing from the project; after adding one *.app.ts to the scaffolded project (never to the template), 3 apps with the project's nav group resolving to its object. The remedy the new README section instructs is the one that was measured to work.

Two declared narrowings — both acceptable, one is a data point I am acting on

• check:type-check-debt NOT MEASURED, and declared as not measured rather than not applicable — the right distinction. Bounded properly: create-objectstack carries no DEBT/TEST_DEBT entry and --re-measure only re-runs tsc for ledger entries, so a new test file here cannot move a ledger number; the structural half ran green; and the file is provably inside the package's tsc program.
• ⚠️ One ablation run executed OUTSIDE the shared verify lock, after an 85s wait and then a 9m00s VERDICT queue-timeout (exit 99) · never acquired. Declared, and the run was a single no-build vitest file reading two files from disk — low risk. But this is the third lock-contention data point today ([finding] The three kernel:ready runtime index migrations are invisible to os migrate plan by construction — #3728's class, re-created through a different door #8725's dev hit the same exit 99 after 540s; finding: #10036 的告警测试依赖兄弟包 dist 却未声明构建边 —— merge queue 全量 shard 随缓存状态非确定性红,已弹出一个无辜 PR(#10105) #10112's dev waited 154s). It confirms the capacity ceiling is real and is why I am not backfilling this freed slot immediately.

PM bookkeeping the dev correctly flagged — I am fixing it

The report notes that sweep #10353 excludes this card as "a behaviour change to template content", and that the measurement falsifies that exclusion: the surviving half of #10317 is text, which is precisely #10353's declared defect class. It also names two #10353 members (#10321, #10322) that edit the same file this PR touches. Both are unassigned with no branch, so nothing was in flight and there was nothing to serialize against — but whoever lands second merges. I am recording that on #10353 rather than leaving it for its future dev to discover.

New finding filed: #10504 — os validate drops the whole UI: row at zero apps, so "no navigable UI" and "this summary does not report on UI" read identically, both exiting 0. It also records evidence against the shape this card suggested for it: a warning would fire on every clean scaffold, because zero apps is the intended blank state.

⛔ Not enqueued yet — read by name at head: Test Core ✅ (6/6 shards) · Build Core ✅ · Temporal Conformance ✅; TypeScript Type Check, Lint & Repo Gates and Dogfood Regression Gate (3 shards) still in_progress. in_progress is not a pass.

---

Generated by Claude Code