fix(plugin-security): explain's record update/delete verdict uses the by-id write path's inputs

[objectstack-fleet]

Fixes #19963

Clause-②: no

What changes

POST /api/v1/security/explain with { object, operation: 'update' | 'delete', recordId } now builds its record verdict from the by-id write path's own inputs. decision.record.visible equals what the by-id PATCH / DELETE does for the same caller. Enforcement admits and refuses exactly what it did before. Only the explain report changes.

Landing: packages/plugins/plugin-security/src/security-plugin.ts (the explainAccessForCaller wiring, plus one helper extracted from the step-2.7 gate), a new test file, the changeset, and three generated rows in scripts/engine-double-contract.pinned.json (see "Surface note").

The two mechanisms, reproduced on origin/main 2c1011b01b before the change

The old bindings, quoted from 2c1011b01b:

• computeLayeredRlsFilter: (sets, o, engineOp, c) => this.computeLayeredRlsFilter(sets as any, o, engineOp, c): no options, so the platform floor stays (M1).
• canEditRecord: (o: string, rid: string, c: any) => sharing.canEdit(o, rid, c): bare context, no __writeScope (M2). canDeleteRecord had the same shape.

One-variable causal legs. Each ran on the unfixed tree through the real SecurityPlugin, the real SharingService and sharing middleware, and the real platform member_default. The object has an unset OWD (private).

leg	the one variable	explain record	by-id PATCH
M1: dept_reporter × edit-shared row	created_by = admin	visible: false, decidedBy: rls	admitted
M1: same principal, row, share and owner	created_by = dept_reporter	visible: true	admitted
M2: hr_reviewer (write depth org, no org_member, so the floor is out of play) × unshared row	bare context	visible: false, decidedBy: sharing	admitted
M2: same call	__writeScope: 'org' stamped on the context	visible: true	admitted

Delete was unmeasured on the card, and it has the same shape. On 2c1011b01b, hr_reviewer × unshared and dept_reporter × owned (owner, not creator) both read visible: false (rls), while the by-id DELETE admitted. This PR covers delete too.

The change

• resolvePreImageFloorDrop(rlsOperation, object, recordId, context, sets, delegated) is step 2.7's floor decision, extracted clause for clause. The floor must be in play, the sharing tri-state verdict must be allow, and the on-behalf-of path is excluded. The gate calls it with the arguments it used before. The whole plugin-security suite stays green (2341 tests), including the Row-level write gate consults neither modifyAllRecords nor sys_record_share.access_level — both declared write-widening mechanisms are inert #5492 / A controlled_by_parent detail's own ownership floor is never droppable, so a cross-creator by-id UPDATE of a child is refused before the master gate — Modify All Data included #8757 / ADR-0055 composition suites that drive this gate.
• Explain's computeLayeredRlsFilter binding passes { dropPlatformOwnershipFloor } from that helper for update / delete. It does not pass masterGateCoversThisWrite: that option vouches that ADR-0055's master gate runs after the filter, and explain runs no master gate. An unanswerable decision keeps the floor standing.
• Explain's canEditRecord / canDeleteRecord bindings stamp __writeScope from resolveWriteScopeForSharing. The stamp always overwrites the context's value, the way step 2.6 overwrites it.
• On-behalf-of contexts keep their previous inputs on both halves (see "Acceptance notes").

Tests: explain-write-verdict-inputs.test.ts (25 cases)

Every cell asserts that record.visible equals the write's outcome. An admitted write must really change the row. A refused write is checked by its ADR-0112 code + status: PERMISSION_DENIED/403 for the pre-image gate, FORBIDDEN/403 for the sharing gate.

• update matrix, private OWD: builtin admin × shared (control); hr_reviewer × shared / owned / unshared; dept_reporter × edit-shared, × owned-not-created. Negative control: dept_reporter × unshared, where both deny.
• delete matrix: hr_reviewer × unshared and dept_reporter × owned admit. dept_reporter × edit-shared (ADR-0111 D3) and × unshared are refused.
• M1: created_by moved only, and the verdict must not move.
• M2: an org writer with no floor is admitted. A __writeScope that the caller brings does not decide the verdict (negative pin).
• OWD control public_read_write: all nine principal × row cells admit on both sides.

Unfixed tree 2c1011b01b: Tests 10 failed | 15 passed (25). Fixed: Tests 25 passed (25).

Ablations

The fix was committed first (81dba2597). Each mutation used scripts/ablation-replace.mjs: the anchor went from 1 hit to 0, and the blob changed on disk. After each restore the blob matched HEAD (be403e07a5ae) and git diff HEAD was empty. A driver trap also re-verified both.

ablation	mutation	result
A1 (M1)	old computeLayeredRlsFilter binding, no options	8 failed / 17 passed: hr_reviewer × 3 update, dept_reporter shared and owned update, 2 delete cells, the M1 created-by-admin leg. The M2 pins stay green.
A2 (M2)	old canEditRecord binding, bare context	5 failed: hr_reviewer × 3 update (decidedBy: sharing), the M2 org-writer pin, and the forged-stamp negative pin (explain true, PATCH refused).
A3 (delete)	old canDeleteRecord binding	1 failed: hr_reviewer × unshared delete.
A4 (every negative pin)	explain forced wide: __writeScope: 'org' stamped and the floor always dropped	5 failed: the update negative control (both cases), delete × edit-shared, delete × unshared, and the forged stamp.

The first A4 attempt was a no-op on its second mutation. The driver read that anchor file after a cd, and ablation-replace refused the empty anchor with exit 2. The re-run with absolute paths is the row above. The test imports ./security-plugin.js from source, so there is no dist/ hop and ablation-dist-preflight does not apply.

Local verification: final head aa9013ffb

• pnpm --filter @objectstack/plugin-security build exit 0. typecheck exit 0: the new test is in tsconfig.test.json's program (--listFiles: 1 hit).
• Full plugin-security vitest run: Test Files 123 passed, Tests 2341 passed. This ran at 81dba2597; aa9013ffb adds only the ledger rows.
• dispatch-gates --commands union at aa9013ffb, reconciled with --ran: 70 derived, 69 ran with exit 0, 1 NOT MEASURED. check:dual-build-cjs-loads exits 3 (PREREQUISITE NOT MET): it needs every workspace package's dist/, and CI's Build Core owns that. Four more roster gates keep their roster under a directory this diff touches: check-changeset-fixed, check:authz-resolver, check:error-code-casing, check:filter-alias-parity. All four exit 0.
• Narrowed eslint (--no-inline-config --format json) over the two changed TS files: 2 files, 0 errors, 0 warnings. Both files are in the population of eslint.config.mjs's **/*.{ts,…} block. That config enables no type-aware linting (no parserOptions.project), so this diff cannot move the verdict of any untouched file.
• origin/main was still 2c1011b01b when this PR opened, so the pre-PR merge was a no-op.

Surface note

scripts/engine-double-contract.pinned.json (+15 lines: the new test file's delete / findOne / update doubles) was written by node scripts/check-engine-double-contract.mjs --write, which is what that gate prescribes for new pinned doubles ("0 seams added or lost"). It lies outside the claim's declared file surface. The step-3.6 post-image block is untouched.

Acceptance notes

• The read-side twin reproduces, and this PR does not change it. Explain binds sharingReadFilter as sharing.buildReadFilter(o, c) with no __readScope. On 2c1011b01b, explain(read) for hr_reviewer (read depth org) answered visible: false, decidedBy: sharing, on all three rows. The same find through the real security and sharing middleware returned each row, because enforcement stamps __readScope: 'org' and the sharing read filter is then null. The dept_reporter (own) control agrees on both sides. This is reported for the seat to file.
• Explain's record path does not model these, before or after this PR. None of them was measured here:
  • the ADR-0090 D10 on-behalf-of legs (the agent-leg depth intersection, and the second gate call as the delegator);
  • ADR-0055's master gate for controlled_by_parent details;
  • the sharing middleware's probeAuthoredRowWrite deferral;
  • the lifecycle verbs transfer / restore / purge, whose Layer 1 is computed with the raw verb rather than the gate's rlsOperation mapping.
• The test double's nested reads are not scoped by plugin-sharing's read filter. That is the same boundary row-write-widener-composition.test.ts records. In every admitted cell, the caller can read the row on the real stack too: through org read depth, an edit share, or ownership.
• Parallel work in the same file. The fold of security: a check-only row-level policy does not gate a bulk update (update(…, { where, multi: true })): the post-image check is skipped as "governed by the using-scoped where", and no using exists to scope it #19950 and [finding] security: a row-level check (declared, or defaulted from using) is never evaluated on an ARRAY insert — engine.insert(object, [rows]), which createManyData calls, stores rows a single-record insert refuses #19964 edits the step-3.6 block of the same file. It had not landed when this PR opened. Whichever of the two lands second merges main and re-runs.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/plugin-security, touching 3 documentable anchor(s).

5 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/permissions/field-level-security.mdx (via SecurityPlugin (symbol, a top-level class))
• content/docs/permissions/index.mdx (via SecurityPlugin (symbol, a top-level class))
• content/docs/permissions/system-context.mdx (via explainAccessForCaller (symbol, a method of class SecurityPlugin))
• content/docs/plugins/packages.mdx (via SecurityPlugin (symbol, a top-level class))
• content/docs/ui/forms.mdx (via SecurityPlugin (symbol, a top-level class))

⛔ 1 release-owned page(s) also name something this change touched. These are read-only:

• content/docs/releases/implementation-status.mdx (via SecurityPlugin (symbol, a top-level class))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• 1 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 60 of 215 client-bound route-ledger rows — the other 155 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 155: 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; 100 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 2c1011b01bc071c545f72f2761647b8d9ab56375 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from fdfb498a4733704a55d4b8654f3da779b8dc628d — the merge of head aa9013ffb9712f779d8fc11bcb4a39b0f1906c49 into base 2c1011b01bc071c545f72f2761647b8d9ab56375, 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 fdfb498a4733704a55d4b8654f3da779b8dc628d && git checkout fdfb498a4733704a55d4b8654f3da779b8dc628d
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 2c1011b01bc071c545f72f2761647b8d9ab56375 aa9013ffb9712f779d8fc11bcb4a39b0f1906c49 && git checkout -B drift-repro 2c1011b01bc071c545f72f2761647b8d9ab56375 && git merge --no-ff aa9013ffb9712f779d8fc11bcb4a39b0f1906c49

node scripts/docs-audit/affected-docs.mjs --json 2c1011b01bc071c545f72f2761647b8d9ab56375

⚠️ 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 2c1011b01bc071c545f72f2761647b8d9ab56375 → pass the list as
args.docs, on the commit named under Which tree this was computed on.