feat(plugin-security): the boot writes no sys_position row — delete the built-in and declared position seeders and the everyone baseline junction writer (ADR-0131 D2, #15204 stage 6a)

[objectstack-fleet]

Part of #15204 (ADR-0131 cutover, stage 6a). This PR deletes the built-in and declared position seeders and the everyone baseline junction writer. The card stays open for the remaining stages.
Clause-②: no

Size: 36 files, +443 / −2,045 (2,488 changed lines) against main at b7cd1af9df. That is under the 3,000 line.

Landing: this PR stays a draft until stage 5 (C9) lands, per order correction 3 of the stage-0 record on #15204. Setup's position pickers still read rows, so on a fresh database they would list no built-in position.

What is deleted

• plugin-security/src/bootstrap-builtin-positions.ts (bootstrapBuiltinRoles, 166 lines)
• plugin-security/src/bootstrap-declared-positions.ts (bootstrapDeclaredPositions, 391 lines), with its test file (410 lines)
• In security-plugin.ts:
  • the seedCatalogPositions, seedCatalogBuiltins and bindBaselineToEveryone closures;
  • their calls in the boot sweep and in the organization-creation hook;
  • the two imports, and the resolveOwnOrganizationRow / seedCtx imports that only they used.
  • seedCatalogForOrganization now runs the one catalog step left (the permission-set seeder, which 6b-2 retires).
• isBuiltinPositionName in builtin-positions.ts. Its only caller was the deleted declared-positions seeder's exclusion.

per-organization-catalog.ts is untouched: 6a never imported it directly, and 6c retires it. bootstrap-platform-admin.ts imported nothing from the deleted modules (measured: no import, two comments only), so there is no seam with 6b-2.

Why no grant moves

Since stage 1 the resolver reads position definitions:

• everyone's baseline is the anchor's declared permissionSets. declareEveryoneBaseline registers it at kernel:ready.
• The set list is this.baselinePermissionSets, which is the app's default set composed with member_default.
• It is filtered by describeHighPrivilegeBits against the same declared-capability context bindBaselineToEveryone used.

So the junction rows this PR stops writing granted nothing. Measured:

• builtin-positions.boot.test.ts boots the real plugin over SQLite in 3 postures × 4 variants, including a new existing database variant that holds the rows the retired seeders wrote.
  • Every principal's grant envelope equals the unchanged golden: platform admin, organization admin, member and agent, fresh and existing.
  • Zero-write census: no sys_position row and no ledger write in any variant. An existing database keeps exactly the rows it had.
  • Positive control: one organization-authored create reads back as exactly one row.
• security-plugin.test.ts: the everyone definition names the app set when its token is declared, and does not when the token is undeclared or names a platform capability. No junction row is written for either token.
• dogfood declared-position-provenance: a real showcase boot writes no position row, and a cold boot of an existing database re-stamps, refreshes and adds nothing.
• The resolver's batch-equivalence golden (resolve-authz-context.batch-equivalence.golden.json) is not touched.

Tests

Every test that imported a deleted module or pinned a deleted row:

file	disposition
bootstrap-declared-positions.test.ts	deleted with its module
audience-anchors.test.ts	3 seeding cases deleted; predicate cases kept
bootstrap-seed-round-trips.test.ts	5 position cases deleted; permission-set cases kept
builtin-positions.boot.test.ts	rewritten: zero-row census + ledger with a positive control; new existing-database variant; grant golden unchanged; built-in-row refusals read on the existing database; the door-authored-rows case deleted
builtin-positions.test.ts	isBuiltinPositionName case deleted with the function
catalog-reference-report.test.ts	precondition rewritten: the built-in copies are staged as existing-database rows
objects/reserved-identity-names.test.ts	platform-seed case rewritten as a direct platform-provenance write; the exemption stays until stage 8 retires the object
per-organization-catalog.test.ts	rewritten onto the permission-set seeder; case 1 adds the zero-position census; case 6 stages the everyone rows
security-plugin.test.ts	the #18535 junction cases rewritten onto the definition; a no-junction-row case added
seed-write-refusal.test.ts	8 seeder-driven cases deleted; the redaction canary (#8682) rewritten at the accumulator
dogfood showcase-declarative-rbac-seeding	positions read through the security catalog; the re-seed case deleted
dogfood declared-position-provenance	rewritten: the gate legs run on staged existing-database rows; the fresh boot writes no row; a cold boot re-stamps, refreshes and adds nothing
dogfood me-apps-and-everyone-baseline, showcase-d7-default-profile	the baseline is read off the everyone definition; no row is written
dogfood showcase-invoice-seed-isolation	the position route rides the contributor declaration; no row or junction row is written
dogfood showcase-permission-zoo	the junction-gate cases stage the row an existing database holds
dogfood position-held-name-setup-create	the precondition expects no seeded row

Docs and ledgers

• Docs: authorization.mdx (the declared-positions bullet and the lifecycle row), positions.mdx, permission-metadata.mdx, system-context.mdx (rows 11b and 23b), and tenant-audit-census.mdx. The census is regenerated (238 → 233 write sites: the 5 deleted writes) and its prose figures are updated.
• Code comments: TSDoc that named the seeders in builtin-positions.ts, reserved-identity-names.ts, position-catalog-refusal.ts, suggested-audience-bindings.ts and bootstrap-platform-admin.ts.
• Ledgers: engine-double-contract.baseline.json, engine-double-contract.pinned.json and objectql-double-limit.baseline.json drop the doubles that went with the deleted cases.
• Not touched: content/docs/releases/ and any packages/spec/src path. 6b-1c, now on main, already rewrote high-privilege.ts's sentence naming bindBaselineToEveryone.

Open question for the seat: the audience-binding suggestion confirm

Settled: the seat answered A in 6106681412 (no change in 6a). The confirm went to decision card #22801, ruled A (6108478638), and stage U2 builds it.

Measured on a real showcase boot of this head, with the app default CLI-wired:

• showcase_member_default is a pending everyone suggestion.
• The database holds no everyone row and no junction row.
• confirmAudienceBindingSuggestion needs the everyone row, so a confirm on a fresh database answers SuggestionStateError ("not seeded yet").

On main, by reading the code (this before-leg was NOT MEASURED in that wiring):

• the boot's junction row made the reconcile record the suggestion as confirmed (observed);
• a confirm of any other pending suggestion wrote a junction row that grants nothing.

In the default wiring both trees measured pending. This is the suggestion-confirm path that release-cut condition (d) of the stage-1 ACCEPT already holds out of every release. No fix is built here (only delete; U2 belongs to its own stage). The report on #15204 carries the options.

Acceptance notes

• Four test comments still describe rows the retired seeder wrote: driver-sql sql-driver-sys-position-organization-unique.test.ts:310, bootstrap-platform-admin-seeded-provenance.test.ts:204, objects/sys-position.organization-unique.test.ts:44 and suggested-audience-bindings-install-path.test.ts:139. Each reads as history of an existing database's rows. Not edited.
• Retired modules are not edited, so their comments still name the deleted seeders: objects/sys-position.object.ts (stage 8), seed-name-lookup.ts (7a), bootstrap-declared-permissions.ts (6b-2) and position-write-through.ts (7b).
• scripts/measure-position-name-fold-census.mjs still models bindBaselineToEveryone as a junction producer. It is a measurement instrument whose question (name-fold vs junction) ADR-0131 retires. Carrier: none.

Gates

Derived with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack at head ac2d982b (merge base b7cd1af9df). Each command was run with its exit code recorded, and the record was reconciled with --ran: 105 derived, 104 run, 1 NOT MEASURED, 0 unrun.

• 104 commands exited 0.
• NOT MEASURED: check-plugin-teardown-shape --self-test exited 3. A shallow clone cannot reach its pinned fixture commit, as on the sibling stages.

Suites on the merged head ac2d982b, each after a dependency-closure build (plugin-auth and the example apps were measured at 68ba543e and not re-run on ac2d982b; CI runs them):

package	files	tests
@objectstack/plugin-security (typecheck + suite)	197	4,005 passed, 45 skipped
@objectstack/plugin-auth	136	2,769 passed, 10 skipped
@objectstack/runtime	352	5,789 passed, 19 skipped
@objectstack/dogfood	246 (+1 skipped)	1,994 passed, 9 skipped
example apps: showcase / crm / todo / embed-objectql	33 / 6 / 7 / 1	409 / 46 / 238 / 2 passed

Before-leg of the deletion: with the deletion in place and only the five importing test files fixed, the full plugin-security suite was red, with 22 failures in 3 files (the boot census, the catalog report's built-in precondition, and the #18535 junction cases). The first dogfood run was red too, with 15 failures in 6 files. Every one of them pinned a row the deleted seeders wrote. CI was not awaited.

Seat edits at landing round 1 (ac2d982b, a merge of main b7cd1af9df plus the census regeneration): the size base, the census figures, the gate head and the suite rows above, from the dev's report 6108751935. Lines 1–2 are unchanged.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/plugin-security, touching 26 documentable anchor(s). ⚠️ 3 changed file(s) yielded no anchor (packages/plugins/plugin-security/src/objects/reserved-identity-names.ts, packages/plugins/plugin-security/src/position-catalog-refusal.ts, packages/plugins/plugin-security/src/suggested-audience-bindings.ts), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files.

27 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json b7cd1af9df7ade29e165e838503cbbef05e4299e.

⛔ 15 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 3 changed file(s) yielded no anchor (packages/plugins/plugin-security/src/objects/reserved-identity-names.ts, packages/plugins/plugin-security/src/position-catalog-refusal.ts, packages/plugins/plugin-security/src/suggested-audience-bindings.ts) — pages documenting those are invisible to this run
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 15 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json b7cd1af9df7ade29e165e838503cbbef05e4299e → packageMentionDocs.

Which tree this was computed on

This run read content/docs from e82fcd5c0b5c730793dbee25a47a4d5d85b04b30 — the merge of head ac2d982b92d7931aefa2d3205dc9d5088e4be044 into base b7cd1af9df7ade29e165e838503cbbef05e4299e, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin e82fcd5c0b5c730793dbee25a47a4d5d85b04b30 && git checkout e82fcd5c0b5c730793dbee25a47a4d5d85b04b30
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin b7cd1af9df7ade29e165e838503cbbef05e4299e ac2d982b92d7931aefa2d3205dc9d5088e4be044 && git checkout -B drift-repro b7cd1af9df7ade29e165e838503cbbef05e4299e && git merge --no-ff ac2d982b92d7931aefa2d3205dc9d5088e4be044

node scripts/docs-audit/affected-docs.mjs --json b7cd1af9df7ade29e165e838503cbbef05e4299e

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs b7cd1af9df7ade29e165e838503cbbef05e4299e → pass the list as
args.docs, on the commit named under Which tree this was computed on.