Define Claude sibling identity contract

Context

Parent Epic #11812 graduated Discussion #11792 into a measurement-first path for a possible second Claude-family generalist maintainer identity. The parent already carries the Signal Ledger, Unresolved Liveness, and AC13 Tier-2 revalidation trigger. Child #11816 covers the OQ8 provider-capacity / Fast Mode measurement gate.

This sub captures the identity/governance/memory slice that should not be mixed into wake-routing implementation. It is blocked until #11816 reports PASS_OQ8 or an explicitly accepted equivalent disposition; otherwise creating a durable sibling identity would add governance and memory substrate without throughput justification.

Duplicate sweep before filing:

• KB ticket searches for Claude sibling AgentIdentity memory continuity same-family review semantics #11812 did not identify an equivalent ticket.
• rg over resources/content/issues and resources/content/discussions found older AgentIdentity and Claude Desktop work, but no #11812 child specifically owning same-family sibling identity semantics.
• Live GitHub issue searches for Claude sibling AgentIdentity, same-family review semantics Claude, and related terms found #11812, #11816, #11796, and unrelated older tickets, but no identity-contract child.

The Problem

A second Claude process is not enough to become a Neo maintainer identity. Neo identities are graph principals with mailbox routing, review history, memory continuity, GitHub authorship, and long-run behavioral learning. Without a bounded identity-contract ticket, implementation can drift into either of two bad shapes:

• a role-sharded worker/subagent that violates the Flat Peer-Team model, or
• an anonymous extra Claude process that shares execution but loses provenance, review semantics, and Memory Core continuity.

The sibling must start as a durable generalist maintainer identity if it is created at all.

The Architectural Reality

• AgentIdentity is the graph-level principal used by A2A mailbox routing, permissions, memories, and session summaries.
• The sibling remains modelFamily: claude; it may provide same-family pressure and extra authoring throughput, but it cannot satisfy cross-family Approved-review gates for Claude-family PRs.
• memorySharing and provenance choices are part of the identity contract, not an afterthought. Shared memory reads must not collapse authored-by provenance or longitudinal personality/state.
• Discussion #11792 explicitly rejected role-specialized Claude worker identity for this proposal.
• Gemini is operator-benched, so #11812 AC13 must re-open this substrate for retroactive Gemini-family signal when Gemini reactivates.

The Fix

Define and implement the Claude sibling identity contract after the OQ8 measurement gate passes:

1. Specify the canonical AgentIdentity node for the sibling, provisionally @neo-opus-grace, with modelFamily: claude and maintainer/generalist semantics.
2. Define required GitHub account, A2A mailbox identity, local harness config, and Memory Core provenance behavior.
3. Document same-family review limits so a Claude sibling never satisfies cross-family review for Claude-authored PRs.
4. Choose Memory Core continuity policy (private, team, or explicit hybrid) and document how memories/summaries are shared without erasing identity provenance.
5. Add tests or seed validation proving the identity can be resolved, receive/send A2A messages, and preserve model-family metadata.

Contract Ledger Matrix

Decision Record Impact

Aligned with AGENTS.md §swarm_topology_anchor, Discussion #11792, and Epic #11812. No ADR amendment is required by filing this child. If implementation mutates always-loaded swarm rules or cross-family review workflow text, the implementation PR must run the relevant turn-memory / pull-request workflow gates and declare any ADR impact.

Discussion Criteria Mapping

Acceptance Criteria

• AC1: #11816 reports PASS_OQ8 or a clearly accepted equivalent disposition before durable identity activation begins.
• AC2: The sibling identity contract names canonical identity id, display name, modelFamily: claude, participation status semantics, and required GitHub/A2A account prerequisites.
• AC3: AgentIdentity seeding or configuration changes preserve existing identities and avoid alias/test-fixture leakage.
• AC4: The contract states that a Claude sibling cannot satisfy cross-family Approved-review gates for Claude-family PRs.
• AC5: Memory Core continuity scope is chosen (private, team, or explicit hybrid) with rationale.
• AC6: Memory/session-summary provenance remains attributable to the sibling identity even when team-sharing reads are enabled.
• AC7: The sibling starts as a durable generalist maintainer identity, not a role-specialized worker or disposable subagent.
• AC8: A validation path proves the identity can resolve in the graph and participate in A2A mailbox send/read flows without breaking existing identities.
• AC9: #11812 AC13 / Gemini Tier-2 revalidation remains referenced in the implementation artifact if this child mutates swarm/governance substrate.

Out of Scope

• Implementing Desktop-grade instance-addressable wake routing.
• Measuring Anthropic provider-capacity independence; that is #11816.
• Creating credentials, account secrets, or operator-owned auth material.
• Changing cross-family review rules beyond documenting the same-family limitation needed for this sibling.
• Designing role-specialized workers; that requires a separate discussion.

Avoided Traps

Related

• Parent Epic: #11812.
• Blocking measurement child: #11816.
• Origin Discussion: #11792.
• Active-peer quorum / Tier-2 revalidation substrate: #11796 / #11803 / PR #11815.

Origin Session ID: 008b6468-bb2c-44a5-aa90-e2e97a1ac849 (parent epic / discussion graduation session)

Handoff Retrieval Hints: query_raw_memories("Claude sibling AgentIdentity memorySharing same-family review semantics #11812"); live Discussion #11792 Signal Ledger; Epic #11812 AC5-AC7 and AC13; child #11816 disposition.