[joemooney]

Operators have a supported Linux setup route separate from conversational onboarding. Fresh clones must run aida init (or aida node acquire) before creating specs. Installation/MCP opt-in, adoption/sibling stores, UUID allocation/HOME authority, cache/local state and transport documentation agree with source. Restored allocation retry/duplicate-ID refusal details, queue list --sync and unpublished-store history checks preserve operational guidance. This adds the seventh curated guide. Automatic pull merge-gating and agreed short-ID promotion require an enrolled HOME clone. Ordinary cache reads use the bounded freshness contract. The setup reference retains the nested-repository and reinitialization --force refusals.

Validation at author head 9016ad0c20371f65975af3070741283fada6e0ac: make check-ci-fast completed all 35 gates successfully at this exact head. Guide output was rebuilt from its source and shared template; this change adds no runtime authority or deployment behavior. A fresh author self-check independently checked the final six ruling items against source and link targets; no scoped findings. Author checks provide no independent approval or merge-hold clearance. Advisor review and bottom-up integration remain required.

[joemooney]

Review: PR-3102 @ 742acec75c0 — Request changes (round 1)

Spec	Verdict	Evidence
TASK-B6B1278890774EF993D2EBC327862470-112	⚠️ PARTIAL	Operator guide and most corrections were checked against source and help. Three route and data-layer inconsistencies remain in touched docs (below).

Acceptance criteria

• ac651625 (verify install/release, init/MCP, adopted/sibling, identity/allocation, data-layer against source/help): ⚠️ mostly verified.
  • Checked: aida init --help (--with-mcp, --agent incl. both/all/PATH detection, --footprint, --minimal, --attach, --store-path); .github/workflows/release.yml and .gitlab-ci.yml are Linux x86_64/arm64 only; scripts/install.sh --version; aida adoption status|inventory, config allocation show, config transport show, upgrade exist; periodic auto-push falls back to manual (store_sync.rs:925).
  • Gap: the data-layer claim at docs/cli/10-storage.md:155 is still stale (finding 2).
• ac7437bc (one consistent route; preserve still-valid detail): ⚠️ findings 1 and 3.
• ac6f8260 (canonical / cache / local state separated; rebuild ≠ restore): ⚠️ verified at operator-setup.md:23-29, 10-storage.md:13-15 and admin-guide.md:238. Contradicted by 10-storage.md:155 (finding 2).
• ac37475e (HOME/WORK and transport reflect shipped code; no future designs as shipped): ✅ operator-setup.md:31-43 matches 07-project-setup.md:387-420 and transport_pull.rs (journaled FF pull). Future hierarchy is explicitly labelled not shipped.
• ac3b6fec (links fixed at source; template mirrors): ✅ AGENTS.md is not a template mirror. No broken relative links in the new and rewritten docs. python3 docs/showcase/build.py reproduces the committed artifacts with zero drift.

Findings (all must be addressed)

1. docs/storage-modes.md:79-83: "New developer setup" still runs git clone; aida list with no aida init / aida node acquire step. This contradicts the corrected route in operator-setup.md:17, README.md:93 and getting-started.md:48.
2. docs/cli/10-storage.md:155 still says "the HEAD-SHA stale-check rebuilds automatically", which contradicts the same chapter's rewritten intro (bounded reads with stale/deferred/busy labels). The "(gitignored, auto-rebuilt)" wording at storage-modes.md:51 and admin-guide.md:231 should be aligned too.
3. docs/multi-node.md "Check identity and routing": the USER-mismatch diagnostic (ls .aida-store/registry/queues/) and the concrete export AIDA_USER=<id> fix were dropped. Their replacement, aida queue list --local --all, only shows the current user's queue, so it cannot reveal items queued under the other name. Restore a per-user inspection (for example --user <other>) and the pinning fix.

CI: PR checks show only merge-hold-gate (pass), because the base is stacked. Locally, make check-ci-fast at this head exited 0.

Recovery: aida rework TASK-B6B1278890774EF993D2EBC327862470-112 --work --resume --reason "PR-3102 review: fix storage-modes new-dev route, 10-storage cache auto-rebuild claim, multi-node USER diagnostic"

[joemooney]

Review: PR-3102 — ⚠️ Request changes (head e1347c3)

Criterion	Status	Evidence
Linux install/release, init/MCP opt-in, adoption/sibling, clone identity, data layer match source/help	✅ verified	aida init --help (--with-mcp, --agent, --footprint, --attach, --store-path); release.yml/.gitlab-ci.yml (Linux-only); install.sh --version; adoption, config allocation/transport show, db status, role status exist
One consistent route across admin/storage/multi-node/multi-hub/setup docs	⚠️ partial	OVERVIEW.md:411, 10-storage.md:54, storage-modes.md:179-180 still contradict the corrected text; 01-getting-started.md:29 drops the shipped nested-repo refusal
Canonical history / cache / local state kept separate; rebuild ≠ recovery	⚠️ partial	operator-setup.md, admin-guide, 10-storage are correct; OVERVIEW.md:411 still says "auto-rebuilt"
HOME/WORK and transport match shipped code	⚠️ partial	transport claims match transport_pull.rs and store_push_pull.rs:4044; 10-storage.md:54 says merge-gate runs inside aida pull, but it runs only for an authoritative HOME clone (store_push_pull.rs:2442-2447)
Broken links fixed at source; template mirrors	⚠️ partial	no mirrors touched (AGENTS.md is hand-maintained); broken links remain in touched docs: admin-guide.md:7 → missing DEVELOPER_GUIDE.md; 07-project-setup.md:185 → bad anchor (should be #aida-add---sanitized-sanitized-intake)

CI: PR CI does not run on this stacked base (ci.yml covers PRs to main only), and merge-hold-gate fails. Ran locally: python3 docs/showcase/build.py --check exits 0, and tests/test_showcase_build.py passes (7 tests).
Untraced criteria: this is a docs-only spec, so the criteria are verified against source/help as shown above. There is no executable behavior to test.

[joemooney]

Final-round author update at 9016ad0c20371f65975af3070741283fada6e0ac.

Scoped automatic pull merge-gating to enrolled HOME; removed unconditional cache rebuild wording; updated node-ID examples and HOME-only agreed-short-ID promotion; fixed administrator/sanitized-intake links; restored nested-repository and reinitialization --force refusals.

All lower-stack fixes are included by ordinary merges; no rebases or force pushes. One fresh author self-check ran its own source/target evidence; no scoped findings. These are author checks, with no independent-review or hold-clearance authority.

Validation: make check-ci-fast completed all 35 gates successfully at this exact head

[joemooney]

Review: PR-3102 @ 9016ad0 — ✅ Approved

Acceptance criterion	Status	Evidence
ac651625 install/init/MCP/adoption/identity/data-layer verified vs source	✅ verified	docs/operator-setup.md; --with-mcp opt-in, --sibling/--attach/--store-path, WORK default/HOME enrollment match init_cmd.rs, allocation_authority.rs; all cited source paths exist
ac7437bc admin/storage/multi-node/multi-hub/setup refs consistent	✅ verified	cross-links resolve (README.md#install, 07-project-setup.md#clone-local-allocation-authority, #aida-config-transport, OVERVIEW.md#bounded-cache-reads); multi-node keeps per-user queue, AIDA_USER, --user, --sync, unpublished-history check, scoped 2026-05-18 incident
ac6f8260 canonical vs cache vs local state; rebuild ≠ restore	✅ verified	operator-setup.md "Keep three kinds of state separate"; 10-storage.md:17; multi-node "Cache and recovery"
ac37475e HOME/WORK + transport reflect shipped code; no future features as shipped	✅ verified	merge-gate in aida pull gated on allocation_authority::resolve(..).authoritative (store_push_pull.rs:2443-2447) matches 10-storage.md:19,54; transport FF pull/sanitized writes exist (transport_pull.rs, transport_profile.rs); hierarchical allocation explicitly labelled not shipped
ac3b6fec broken links fixed at source; mirrors via master	✅ verified	admin-guide DEVELOPER_GUIDE.md link → agents/aida-repository-guide.md; sanitized-intake anchor resolves to 07-project-setup.md:555; no touched file is a template mirror

Advisor round-3 items (1–6): all fixed at this head.
CI: only merge-hold-gate reports on this stacked PR (governance hold, not a content defect). Docs-only change; the author reports make check-ci-fast passing all 35 gates at this head.
Findings: none.

[joemooney]

Included in PR-3103, which merged the approved stack 107-113 to main as one PR (advisor 11:35).