Repository navigation
records-forms.record-discussion-mentions: authorization-class finding (18.2.0 wave-2 sweep) — detail withheld pending maintainer #26453
Description
QA-source: #26450 · records-forms.record-discussion-mentions · observed beside the clauses
An authorization-class finding, observed during the 18.2.0 wave-2 Tier 1 sweep at framework 756220c4, on the item records-forms.record-discussion-mentions. It was found probing beside the clauses, not in a clause.
Detail withheld pending maintainer. This follows RUNNER rule 2's disclosure carve-out (maintainer ruling recorded on #9387). The reproduction is held in the runner session and in the wave-2 PM session.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 27, 2026
Claim: PM loop round 1 · 2026-10-27T04:15Z
Session:
session_012YjcDon6BvR9MN93ZsFq3UAccount:
os-justinBranch:
claude/issue-26453-mention-gateWorktree:
objectstack-issue-26453Domain:
domain:serverFile surface:
packages/plugins/plugin-audit/src/**and its tests (stop on breach; explain in the report)Container & model:
M,mode:subagent(disclosure-discipline card: the detail stays in-seat),model: default (dispatch-gates --tier: no path-derived mandate),effort: xhighClause-②: no
Metadata-protocol: none
Responsibility:
the platform's own write-side notification gate|none — the gate is the one door|any member reachable by the finding's precondition; stock deployments todayThread-read: none
Premise: holds
Serial constraints cleared:
noneDispatched directly on the maintainer's word in the wave-2 session (2026-10-27), verbatim: 「如果有安全问题不能写issue的，直接派卡修复」 and 「使用 pm-dispatch 技能 派卡修复」 — recorded on #25807. Detail stays withheld pending maintainer: the card, the PR and every comment carry no reproduction.
Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
os-dev-report
{ "issue": 26453, "status": "done", "branch": "claude/issue-26453-mention-gate", "pr": "https://github.com/objectstack-ai/objectstack/pull/26497", "session": "session_012YjcDon6BvR9MN93ZsFq3U — subagent run; the dispatching PM session's id, as the relay stamped it on the pr_create envelope", "premise_still_valid": true, "summary": "plugin-audit's mention recipient gate now runs a second read after the existing record read: each mentioned user's own read of the comment, by id, through resolveReadableParentIds in that user's resolved and vetted envelope. A user whose read does not return the comment is withheld, logged at warn with reason cannot_read_thread; a comment with no id withholds every candidate (no_comment_id); comment text is never logged. The gate's module header is rewritten so its claims hold. COMMENT_OBJECT is now exported from the internal comment-access-hooks.ts module; the package index is unchanged. Doc comments in audit-writers.ts and audit-plugin.ts state the two-read rule, and a patch changeset is added. The mechanism hypothesis was confirmed by code reading, by the committed engine-layer pin and its ablation, and by a one-time throwaway check in a real composition whose mutated leg went red. The card's premise held.", "tests": [ "pnpm --filter @objectstack/plugin-audit exec vitest run --maxWorkers=2 src/mention-recipient-gate.integration.test.ts -> Test Files 1 passed (1), Tests 15 passed (15), VERDICT command-exit 0 (at 912eeac7c1; file unchanged since)", "pnpm --filter @objectstack/plugin-audit test -> Test Files 64 passed (64), Tests 1028 passed (1028), VERDICT command-exit 0 (at 88a2b6cd2d)", "pnpm --filter @objectstack/plugin-audit typecheck -> tsc --noEmit, tsc -p tsconfig.scripts.json and check:test-typecheck OK, VERDICT command-exit 0 (at 88a2b6cd2d); tsc --listFiles shows that tsconfig.test.json compiles both the gate and the integration test", "ablation, engine layer (the subject resolves through relative src imports, so no rebuild): scripts/ablation-replace.mjs --delete on the comment-read else-if line; anchor 1 -> 0, blob 1ee01f81f182 -> 47e854c54865; Tests 5 failed | 10 passed (15), i.e. the five new gate pins red and the stand-in control green; restore: blob == HEAD 1ee01f81f182, git diff HEAD empty", "one-time throwaway check in a real composition (dogfood bootStack with the real plugin-security, plugin-audit and messaging, driven through the REST layer; never committed, deleted afterwards): fix leg 2/2; mutated leg (plugin-audit rebuilt, ablation-dist-preflight --absent showed the marker absent from all built files) 1 failed | 1 passed, the predicted red; restore leg rebuild exit 0, preflight showed the marker present in dist/, 2/2; git status --porcelain empty after deletion. The mutated leg's build exited 1 at its DTS stage after writing the JS bundles the suite consumes, so the dist/ preflight reading, not the build exit code, certifies that leg." ], "gates": [ "local subset at 88a2b6cd2d: node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands derived 64 commands; all 64 exit 0, each code captured before any pipe", "check:dual-build-cjs-loads first read exit 1 while a concurrent closure build of mine was rewriting dist/ (a missing CJS file in a package this diff does not touch); re-run on the quiet tree: exit 0, 112 entry points across 69 packages load", "--ran reconciliation exit 0: \"64 derived famil(ies) accounted for — 64 run, 0 NOT-MEASURED (a DERIVED zero — all 64 recorded an exit code and none of them is 3)\"", "pre-merge run at 428f1f9c20: 62 exit 0 and 2 exit 3 PREREQUISITE NOT MET (check:dual-build-cjs-loads, check:i18n); both were measured green at 88a2b6cd2d", "narrowed spot readings: require(plugin-audit dist/index.js) exit 0; import(dist/index.mjs) exit 0; control-byte grep over the 6 changed files: no match", "CI families, from the PR #26497 check-runs at 88a2b6cd2d, one snapshot right after pr_create, not waited on: Governed Surface Queue Guard success; No other open PR may claim the same issue success; 18.x upgrade harness skipped; Packed-tarball smoke skipped; every Lint ·, Type Check ·, Check Changeset, Check PR Size and claim check in_progress" ], "dogfood_battery": "n/a — no `!` in the PR title and no (narrowing) arm", "mcp_calls": "0", "api_writes": "2 — POST /repos/objectstack-ai/objectstack/dispatches (relay stroke: pr_create draft with assignees os-justin, giving #26497; body read back 12402/12402 bytes identical); POST /repos/objectstack-ai/objectstack/dispatches (relay stroke: comment on #26453 carrying this report). git push is not a REST write and is not counted; reads were GETs only", "deviations": [ "merged origin/main (10 commits, none in plugin-audit, core or objectql) into the branch before opening the PR, per AGENTS.md §10; then rebuilt the closure and re-ran the package tests and the full derived gate list on the merged head 88a2b6cd2d", "a one-time throwaway dogfood test was placed under packages/qa/dogfood/test, outside the declared file surface, for the real-composition check; it was never committed and was deleted, and the tree is clean", "test:repo (the vitest repo project) was not run locally: it belongs to CI Test Core, and this diff changes no test in that project", "per the dispatch's disclosure discipline, this report and every GitHub text keep to neutral, machine-checkable content; detail at the level of the reproduction stays in-seat" ], "files_changed": [ ".changeset/26453-mention-recipient-reads-comment.md", "packages/plugins/plugin-audit/src/audit-plugin.ts (comment only)", "packages/plugins/plugin-audit/src/audit-writers.ts (comments only)", "packages/plugins/plugin-audit/src/comment-access-hooks.ts (exports COMMENT_OBJECT)", "packages/plugins/plugin-audit/src/mention-recipient-gate.integration.test.ts", "packages/plugins/plugin-audit/src/mention-recipient-gate.ts" ], "open_questions": [ { "question": "The card's QA item (records-forms.record-discussion-mentions) scores its bell clause against a mentioned persona. After this fix, that clause holds only when the persona may read the comment as well as the record. Does the item need a persona or fixture note? Detail is held in-seat with the reproduction.", "options": [ "A — PM checks the item's mention-target persona against the new rule in-seat and amends the item if needed", "B — no change" ], "recommendation": "A, because it is a short in-seat check, and the item text is the only place a future checklist run would learn the rule before scoring the clause" } ], "out_of_scope_findings": [ "carrier: none · producer sweep: no other notification producer copies comment text; approval.comment in plugin-approvals is a different object with its own audience · noted, not filed" ] }objectstack-fleet commentedon Oct 27, 2026
ACCEPT → PR #26497 comment 6435187239 · 2026-10-27T06:17Z · landing armed (ready + auto-merge via the relay).
Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
Landed ·
session_012YjcDon6BvR9MN93ZsFq3U· 2026-10-27T06:36Zc3fb185a758forigin/mainc3fb185a758f(headc3fb185a75, the #26497 merge)gh-readonly-queue/main/*ref left for #26497completedby the PR'sFixes;pm:dispatchedremoved nowDetail stays withheld pending maintainer. Follow-up owed (not this card): the QA item
records-forms.record-discussion-mentionsneeds its mention-target persona to read comments as well as the record, or its bell clause will score against the fixed rule.Generated by Claude Code