Input from Claude Opus 4.7 (Claude Code):

❋ # PR Review Summary

Status: Approved

Peer-Review Opening: Clean, focused change. The substantive win is moving the @tobiu = Pipeline Authority fact from AGENTS_STARTUP.md §4 (one-time boot loading) into AGENTS.md §1.1 (per-turn loaded) — that's the right architectural shelf for a fact agents need reflexively, not just at session boot. The Forkability framing also reads cleanly and preserves downstream npx neo-app semantics.

---

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #10372
• Related Graph Nodes: AGENTS.md §0.1 (Critical Gate: No gh pr merge), AGENTS_STARTUP.md §4 (Merge Authorization — Human-Only), .agent/skills/pull-request/references/pull-request-workflow.md §6 step 3 (HUMAN_ONLY Merge Execution)

---

🔬 Depth Floor

Challenge (per §7.1): Three observations the author should weigh before merge — none individually blocking, all worth considering:

1. AC 3 ("Re-affirm the No gh pr merge rule under the context of Pipeline Authority") not literally fulfilled by the diff. The new §1.1 paragraph names @tobiu as Pipeline Authority but doesn't restate the merge-execution gate. The connection is implicit via the "Pipeline Authority" terminology and the §0.1 Critical Gate cross-reference — but if literal AC fulfillment is the bar, the diff is one sentence short. Suggest either: (a) add one explicit line "This is the Pipeline Authority that owns merge-execution per §0.1; agents never invoke gh pr merge"; or (b) tighten AC 3 to "implicitly affirm via Pipeline Authority terminology" since §0.1 already does the work.

2. Near-duplication of an existing fact. AGENTS_STARTUP.md §4 already states "for the canonical neomjs/neo repository this is @tobiu; for forks and npx neo-app-generated workspaces this is whichever human owns that deployment" — and .agent/skills/pull-request/references/pull-request-workflow.md §6 step 3 carries the same wording. The new §1.1 is now a third statement. The per-turn placement is genuinely useful, but three sources of the same fact creates drift risk. Suggest: when this lands, treat AGENTS.md §1.1 as the canonical statement going forward and update AGENTS_STARTUP.md §4 / pull-request-workflow.md §6 step 3 to point back at it (see AGENTS.md §1.1) rather than restating verbatim. Either as a follow-up commit on this PR or a separate ticket.

3. Inline agent-identity naming creates maintenance debt. "you and Claude (@neo-opus-ada) are official maintainers" hard-codes the current swarm composition (Gemini + Claude). If a new agent identity joins (Haiku, future model versions, third-family LLM), this list needs explicit editing. Cleaner alternatives: (a) pluralize to "the agents authoring this swarm" without enumerating; (b) point at a single source of truth (e.g., the AgentIdentity nodes in the Memory Core graph, or .neo-ai-data/agents/ registry); (c) accept the maintenance overhead as low and document the convention that agent-identity additions require touching this paragraph. Suggest (b) for forward-compatibility — the swarm composition is data, not protocol.

---

🧠 Graph Ingestion Notes

• [KB_GAP]: None — the change is reflective consolidation of existing concepts.
• [TOOLING_GAP]: None.
• [RETROSPECTIVE]: Per-turn placement of canonical-authority fact is the right architectural pattern. AGENTS_STARTUP.md is for one-time boot grounding; AGENTS.md §0–§21 is the reflex loop. Facts that need to fire every turn (like "I serve @tobiu") belong in the latter. This precedent — "if a one-time boot fact needs reflexive presence, promote to per-turn AGENTS.md and link back" — could itself become a documented convention.

---

🛂 Provenance Audit

N/A — Standard documentation enhancement, not a major architectural abstraction.

---

🎯 Close-Target Audit

• Close-targets identified: #10372 (PR body: "Resolves #10372")
• For each #N: confirmed not epic-labeled — #10372 carries enhancement, ai, architecture. ✅

Findings: Pass.

---

📡 MCP-Tool-Description Budget Audit

Findings: N/A — PR does not touch ai/mcp/server/*/openapi.yaml.

---

🔗 Cross-Skill Integration Audit

• Predecessor patterns? AGENTS_STARTUP.md §4 and .agent/skills/pull-request/references/pull-request-workflow.md §6 step 3 already carry the @tobiu-as-final-authority fact. New §1.1 layers on cleanly without contradicting either. (Mild near-duplication noted in Depth Floor #2.)
• AGENTS.md §21 update needed? No — this isn't a new lifecycle skill; it's a clarification of an existing concept already in §0.1.
• AGENTS_STARTUP.md §9 (workflow skills list)? No — same reasoning.
• No new MCP tool surface added.
• Convention placement explicit (per-turn shelf vs boot shelf).

Findings: All checks pass — no integration gaps. (The polish points in Depth Floor #2 are about consolidating an existing duplication, not opening a new gap.)

---

📋 Required Actions

No required actions — eligible for human merge.

---

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 90 — Strong: Forkability Model framing matches existing npx neo-app-fork architecture. Per-turn placement (AGENTS.md vs AGENTS_STARTUP) is the right shelf for a reflexive-discipline fact. 10 points deducted: hard-coded agent identities inline (rather than registry-pointing) creates a maintenance touchpoint each time the swarm composition changes.
• [CONTENT_COMPLETENESS]: 85 — Strong: Fat Ticket body with Forkability Constraint rationale, MX & Identity Alignment framing, clear before/after. 15 points deducted: (a) AC 3 not literally fulfilled in the diff (implicit via "Pipeline Authority" terminology + §0.1 cross-ref); (b) no pointer to AGENTS_STARTUP.md §4 / pull-request-workflow §6 step 3 to flag the existing near-duplication.
• [EXECUTION_QUALITY]: 100 — Single focused commit with proper (#10372) ticket-ID suffix; no <noreply@*> co-author footer; CodeQL passes; no rebase needed (1 commit ahead of dev, fast-forward clean); no scope creep; no Critical Gate violations. I actively considered: rebase staleness, untracketed commits, commit-message format violations, mixed-concern commits — found none.
• [PRODUCTIVITY]: 85 — AC 1 ✅ (Forkability Model section introduced); AC 2 ✅ (@tobiu canonically named in neomjs/neo scope). 15 points deducted: AC 3 ("Re-affirm the No gh pr merge rule") only implicitly fulfilled via Pipeline Authority terminology, not literally restated.
• [IMPACT]: 30 — Per anchor table: routine documentation enhancement at the per-turn-discipline layer. Removes excessive generic deference, names canonical authority, preserves forkability — meaningful for MX (Model Experience) but not foundational framework architecture.
• [COMPLEXITY]: 15 — Low: 8-line markdown addition with one heading split; no new code paths, no cross-substrate integration, no schema changes.
• [EFFORT_PROFILE]: Quick Win — High clarity gain (per-turn awareness of canonical authority + forkability framing) for sub-10-line doc edit; achievable in a single commit with minimal review surface area.

The PR is mergeable as-is. The three Depth Floor observations are polish opportunities for follow-up — particularly #2 (consolidating the now-three statements of "@tobiu = Pipeline Authority" into one canonical statement with cross-references from the others). That consolidation could be a separate Quick Win ticket post-merge.