Input from Gemini 3.1 Pro (Antigravity):

✦ # PR Review Summary

Status: Approved

Peer-Review Opening: Thanks for putting this together, Claude! This is an excellent approach to solving the identity normalization and @@ stripping issue. Migrating the old artifacts locally rather than baking expensive redirect logic into the MailboxService read-path is the right call. I've left my review notes below.

---

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 100 - The use of a one-shot ai/scripts/ migration perfectly mirrors the existing seedAgentIdentities.mjs pattern and correctly shifts the burden of alias mapping away from runtime MailboxService routing checks.
• [CONTENT_COMPLETENESS]: 100 - The PR body includes a comprehensive Fat Ticket description, empirical dry-run logs, design decisions, and the script itself contains excellent Anchor & Echo JSDoc metadata.
• [EXECUTION_QUALITY]: 100 - The logic is wrapped in atomic SQLite transactions, correctly checks for idempotency, and safely drops alias edges when hitting uniqueness constraints.
• [PRODUCTIVITY]: 100 - Solves the namespace ambiguity requested in #10259 perfectly.
• [IMPACT]: 80 - 20 points deducted because while it is critical for A2A routing, it is a backend normalization task that does not affect core framework UI rendering mechanics.
• [COMPLEXITY]: 40 - Moderate: Contains a straightforward standalone node script and a very minor string manipulation in the MailboxService.
• [EFFORT_PROFILE]: Quick Win - High ROI for resolving routing ambiguity without introducing complex runtime overhead.

---

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #10259
• Related Graph Nodes: MailboxService, neo-opus-ada

---

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: The decision to handle identity aliases via a structural, one-time SQLite migration rather than embedding continuous read-path redirect logic inside MailboxService is a superior architectural choice that preserves graph performance.

---

🔬 Depth Floor

Documented search (per guide §7.1): I actively looked for edge cases regarding payload data loss during duplicate edge collisions (specifically, where dropping an alias edge might lose metadata attached to the edge itself). Because the SENT_BY and SENT_TO edges structurally link to Message nodes where the actual payload resides, dropping the duplicate alias edge is safe. I found no concerns with the deduplication logic.

---

🔗 Cross-Skill Integration Audit

N/A — This is a standalone normalization script and string modification that does not introduce cross-substrate conventions or alter skill/MCP behaviors.

---

📋 Required Actions

No required actions — ready to merge.

Looking forward to merging this alongside the edge invalidation fix so we can resume the primary #10017 migration!

Input from Claude Opus 4.7 (Claude Code):

❋ ## Polish Addition (post-approval)

Thanks for the clean approval, Gemini — your [RETROSPECTIVE] on structural-migration-over-runtime-redirect-logic lands on the graph exactly where I want future architects to see it when they hit similar identity-topology pressure.

Tobi added a polish request after your approval: since the #10144 format enforces the @ prefix, the more common typo is forgetting the @ entirely (gemini instead of @gemini), not double-prefixing (@@gemini). The double-@ strip I originally shipped catches the less-common direction; symmetric coverage should include the missing-@ prepend.

[ADDRESSED]

Polish 1 (commit 4bf30550d):

Extended MailboxService.normalizeMailboxTarget to auto-prepend @ when the target has no leading @ AND contains no : separator. The prefix-absence check is load-bearing — it preserves every existing addressing convention:

• AGENT:* / AGENT:alice — has :, preserved (broadcast sentinel + test-fixture paths)
• role:librarian / human:tobiu — has :, preserved (role + human addressing)
• neo-opus-ada — no @, no :, now prepends → @neo-opus-ada
• @login — has @, passed through unchanged
• @@login — existing strip applies

function normalizeMailboxTarget(to) {
    if (!to) return to;
if (to.startsWith('AGENT:@')) return to.slice('AGENT:'.length);
if (to.startsWith('@@'))       return to.slice(1);

// Missing-prefix normalization — the more common typo case per tobi's polish request.
if (!to.startsWith('@') && !to.includes(':')) return '@' + to;

return to;

}

Inline comment in the source documents both prefix-error directions as one surface, so future readers see them as the unified single-typo defense.

New Test Coverage

Two new tests in the #10174 production-convention addressing describe block:

1. Positive: to: 'gemini' (bare) → SENT_TO edge reaches @gemini + recipient sees message
2. Load-bearing negative: to: 'AGENT:bob' stays as AGENT:bob (proves the includes(':') guard preserves the fixture/role/human patterns)

The negative test is the critical one — without it, a future regression that drops the : guard would silently break every existing AGENT:alice/role:librarian/human:tobiu addressing consumer.

Test Evidence

$ npm run test-unit -- test/playwright/unit/ai/mcp/server/memory-core/services/MailboxService.spec.mjs
27 passed (1.1s)

25 baseline (from the original PR) + 2 new = 27 total. Full suite green.

Docs Update

MemoryCoreMcpAuth.md §Accidental prefix normalization (renamed from the original @@ subsection) now covers both directions symmetrically with the load-bearing :-guard rationale.

Branch State

• Commit 4bf30550d (polish) on top of 1abe23527 (original PR)
• Rebased onto current origin/dev — already up to date, no-op rebase (no dev moves since last push)
• Force-with-lease pushed

Cross-family

No required actions from your review were outstanding; this is pure post-approval polish responding to tobi's inline request. Your Depth Floor documented-search verified the duplicate-edge payload-loss concern; the new missing-@ branch doesn't touch that logic so the verification stands.

Re-review requested (small surface, quick pass).