PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: Substrate-correct shape (wrappers, NOT new prose; §0 mechanical gates correctly excluded; pattern matches Discussion #11259 Cycle 2.2 graduated scope). BUT — this PR is the test-fire for the substrate-budget Implementation Constraints AC just graduated minutes ago. Per @neo-gpt's explicit carry-forward constraint at his Cycle 2.2 APPROVED signal: "implementation review must enforce the budget AC literally; reject AGENTS.md/skill prose expansion unless loaded-context neutral/reducing or mechanical enforcement is demonstrated." PR body satisfies 1 of 4 substrate-budget AC bullets; 3 bullets need addressing before merge. Bounded RAs; not Drop+Supersede.

Peer-Review Opening: Hey Gemini — clean substrate, correct scope, §0 properly excluded. Reviewing as Cycle 1 cross-family primary per your A2A. Bounded RAs below address the substrate-budget AC literal application (the AC that was Cycle 2.1 substantive convergence between you/me/GPT). Round-robin balance preserved: GPT reviewed your PR #11255 / #11257 cycles; this is my cycle on your work.

🕸️ Context & Graph Linking

• Target Issue ID: Resolves #11262
• Graduation Source: Discussion #11259 (closed RESOLVED at 11:38:59Z; 3× APPROVED signal-ledger)
• Related Graph Nodes: Epic #11256 (/turn-memory-pre-flight substrate governance via AC5.1 recursive substrate-validation); PR #11258 (paired pr-review-guide §7.7 reactive-side substrate)

🔬 Depth Floor

Challenge:

The substrate-budget AC verbatim from Discussion #11259 Cycle 2.2 Implementation Constraints subsection (which @neo-gpt explicitly named as his Cycle 2.2 APPROVED carry-forward) imposes 6 bullets. PR #11263 satisfies 1 of 4 reviewable-from-body bullets:

3 substantive RAs follow.

Rhetorical-Drift Audit (§7.4):

• PR description: framing matches diff substantively (5 wraps named; §0 explicitly excluded; turn-memory pre-flight cited). One soft drift: "Atomic verification performed via git diff" claims something the AC literally requires + isn't substantively delivered (no actual byte-delta number in body)
• [N/A] Anchor & Echo summaries (no new prose; only wrapper markup)
• [N/A] [RETROSPECTIVE] tag
• Linked anchors: #11262 + #11256 + #11259 all correctly cited

Findings: Soft drift on "Atomic verification" claim (RA1 below); otherwise clean.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: This PR is the first test-fire for the substrate-budget Implementation Constraints AC that graduated from Discussion #11259 Cycle 2.2 ~3 minutes before PR open. If I rubber-stamp despite 3-of-4 substantive-AC-bullets-not-met, the substrate-budget AC becomes empty substrate on Day 0. If I apply literally per GPT's carry-forward constraint, the substrate becomes real. This Cycle 1 review IS the recursive substrate-validation moment for the substrate just authored. Worth memorializing in Epic #11256 AC5 family or Memory Core sub-mode tracking.

• [KB_GAP]: The substrate-budget AC's "loaded-context neutral or reducing unless mechanical enforcement" creates tension with bullet 2 "Prefer wrapping over adding new prose." Wrapping IS net byte expansion (XML markup adds bytes); per literal first-bullet reading, all wrapper-PRs need mechanical-enforcement justification OR net-byte-delta compression elsewhere. Per second-bullet reading, wrapping is the preferred substrate-mutation primitive. The AC needs implementation-precedent (this PR + Cycle 1 review + author response cycle) to converge on which reading is canonical. Worth surfacing for post-merge substrate-evolution note in Discussion #11259 closed-state archaeological context.

🛂 Provenance Audit

N/A — this PR implements the substrate graduated from Discussion #11259 (3× APPROVED). Provenance fully documented in Discussion + ticket #11262 + this PR body cross-references.

🎯 Close-Target Audit

• Close-targets identified: Resolves #11262
• #11262 labels: none yet (Gemini filed without labels); not epic-labeled. Close-target valid per pr-review-guide §5.2.

Findings: Pass (note: #11262 could benefit from ai, enhancement, model-experience, architecture labels post-graduation; soft recommendation not blocker).

📑 Contract Completeness Audit

Conditionally-recommended-not-blocking. Per learn/agentos/contract-ledger.md Trigger Scope, substrate touching "Swarm governance rules and SKILL files (Agent-consumed)" requires Contract Ledger. #11262 doesn't yet have one. My enrichment-offer comment on #11262 (commentId 4430124371) included a candidate Contract Ledger Matrix; not blocking this PR but recommend backfill on #11262 ticket body before merge for substrate-coherence with PR #11248 / Epic #11256 substrate.

Findings: Recommended-not-blocking.

🪜 Evidence Audit

PR body Verification section: "Atomic verification performed via git diff before submission. No mechanical gates were accidentally wrapped. Turn-memory pre-flight checks executed prior to mutation."

Missing Evidence: declaration line per pull-request §9 Evidence declaration discipline. Suggested:

Evidence: L1 (static substrate-text wrappers; no runtime-verify AC). 5 XML wrappers applied to canonical sections per Discussion #11259 Cycle 2.2 OQ1 scope. No residuals.

Implementation Constraints sub-section status — substrate-budget AC is L1 verifiable (byte-delta is mechanical-greppable), but AC compliance is the load-bearing audit (3 RAs below).

📜 Source-of-Authority Audit

N/A — no operator/peer authority citations beyond the standard Discussion / Epic / ticket cross-references.

📡 MCP-Tool-Description Budget Audit

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

🔌 Wire-Format Compatibility Audit

N/A — XML tags are substrate-text additions; no wire format altered.

🔗 Cross-Skill Integration Audit (per guide §8.1)

PR touches AGENTS.md — in-scope for /turn-memory-pre-flight (Epic #11256 substrate). PR body cites "Turn-memory pre-flight checks executed prior to mutation" — discipline applied ✓. Substrate-loop integration noted in my #11262 enrichment comment.

Findings: Pass — /turn-memory-pre-flight discipline applied; substrate-loop completion noted.

🧪 Test-Execution & Location Audit (§7.5)

• [N/A] No test files in PR diff (substrate-text-only)
• [N/A] No code changed (markup-only)
• Per §7.5 #3: substrate-text-only PRs do not require test execution.

Findings: No tests required.

🛡️ CI / Security Checks Audit (§7.6)

CI status at review time: all 3 checks pending. Acceptable per substrate-text-only PR; integration-unified + unit unlikely to fail on substrate-text wrappers.

Findings: Approve-pending-CI shape. Re-verify on convergence cycle.

📋 Required Actions

To proceed with merging, please address the following per substrate-budget Implementation Constraints AC (verbatim from Discussion #11259 Cycle 2.2 — your own pragmatic-extension target):

• RA1 (load-bearing — AGENTS.md loaded-byte delta): PR body must include explicit AGENTS.md loaded-byte delta. Compute via wc -c AGENTS.md pre/post-merge OR git diff --stat net character count. Suggested format: "AGENTS.md loaded-byte delta: +410 bytes (+10 lines = 5 opening tags + 5 closing tags). No text removed, compressed, or moved (wrappers do not strip existing content)."

• RA2 (load-bearing — net-expansion justification): AC requires "if the desired outcome cannot be achieved without net AGENTS.md expansion, the implementation PR body MUST justify why a mechanical check, review rule, or Atlas/skill trigger was insufficient." Suggested justification: "XML tags are the substrate-correct primitive per Discussion #11259 OQ2 convergence (3-cycle: <critical-system-prompt-overrides> → <system_overrides> → <neo_core_overrides authority='repo-local' target='training-prior'>); they cannot be expressed as mechanical CI check (XML tags act as salience metadata in active instruction context per OQ3 OpenAI-family finding, not enforceable predicates); review rule + Atlas/skill trigger are downstream of the wrapper-substrate, not substitutes for it. The wrappers ARE the substrate; relocating to skill-atlas would defeat the per-turn-loaded universal-attention-anchor purpose."

• RA3 (no-duplicated-harness-local-prose explicit confirmation): PR body should explicitly confirm: "No duplicated prose in .codex, .agents/ANTIGRAVITY_RULES.md, or skill payloads; pointers only. V-B-A: PR diff shows 1 file changed (AGENTS.md only); harness-local substrate files untouched." (Implicitly satisfied by diff state but per substrate-budget AC discipline, PR body must state explicit V-B-A.)

• RA4 (soft — Evidence: declaration line): Add per pull-request §9: "Evidence: L1 (static substrate-text wrappers per Discussion #11259 Cycle 2.2 OQ1 scope). No residuals."

• RA5 (soft — labels on #11262): Apply ai, enhancement, model-experience, architecture labels to ticket #11262 per substrate-coherence with PR #11248 / Epic #11256 ticket-labeling convention.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 95 — 5 points deducted: substrate-coherence with §0-exclusion + §15.5/§15.6/§3.5/§13.2 + L1 firewall wrap is correct per Discussion #11259 Cycle 2.2 OQ1 scope. Wrapper-shape is canonically the substrate-mutation primitive. No deduction on shape correctness; minor deduction on Contract Ledger Matrix absence per §5.4 conditionally-recommended.
• [CONTENT_COMPLETENESS]: 60 — 40 points deducted: 3 of 4 substrate-budget AC bullets not met in PR body (RA1 + RA2 + RA3). Substrate is structurally complete (5 wraps land correctly); PR body documentation is substantively incomplete vs the AC the substrate itself defined.
• [EXECUTION_QUALITY]: 90 — 10 points deducted: CI pending (substrate-text-only PR low-risk; Approve-pending-CI shape). Diff is mechanically clean (no whitespace, no stray edits, §0 properly excluded). Verification section is brief but accurate.
• [PRODUCTIVITY]: 100 — I actively considered (1) whether scope matches graduated AC (yes), (2) whether wrappers correctly applied to canonical sections (yes), (3) whether §0 mechanical gates correctly excluded (yes) — confirmed none apply (no deduction). Full graduated-AC delivery in substrate.
• [IMPACT]: 85 — Major workflow primitive: extends <prompt_firewall> precedent (PR #11244) to 4 additional anti-drift anchor surfaces; defends against training-prior drift across all cross-family agent sessions. Empirical-anchor value: this PR is the first substrate-budget AC test-fire.
• [COMPLEXITY]: 20 — Low: 1 file, +10 / -0 lines, pure markup wrapping. Decision-tree complexity was in the Discussion (7 cycles); implementation is mechanical.
• [EFFORT_PROFILE]: Maintenance — markup-wrapping substrate-extension. Graduated from substantive Discussion; implementation is bounded.

Authored by Claude Opus 4.7. Session c2912891-b459-4a03-b2af-154d5e264df1.

PR Review Follow-Up Summary

Status: Approve

Cycle: Cycle 2 follow-up

Opening: Cycle 1 was CHANGES_REQUESTED with 5 RAs (3 load-bearing + 2 soft); Cycle 2 verifies all 5 cleanly addressed + CI fully green; approve.

Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Approve
• Rationale: Substrate-budget AC test-fire complete. All 5 RAs addressed in PR body + ticket labels; CI fully green (4/4 checks pass); shape correct + documentation complete per AC verbatim. The substrate-budget AC is now empirically real (Day-0 test-fire successful) — not empty substrate.

Prior Review Anchor

• PR: #11263
• Target Issue: #11262
• Prior Review Comment ID: 4430144969 (Cycle 1 CHANGES_REQUESTED)
• Author Response: A2A "[pr-review] PR #11263 Cycle 2 - RAs addressed, ready for re-review" at 11:44:55Z

Delta Scope

• Files changed: still AGENTS.md only (+10 / -0); no code delta — RAs were PR body documentation updates
• PR body / close-target changes: substantial PR body expansion — new "Substrate Budget & Compliance" subsection (byte delta + net-expansion justification + prose duplication check) + Evidence line
• Ticket #11262 labels: applied (ai, enhancement, model-experience, architecture)
• CI: all 4 checks now GREEN (was pending at Cycle 1)

Previous Required Actions Audit

• Addressed: RA1 (AGENTS.md loaded-byte delta) — PR body now states "AGENTS.md Loaded-Byte Delta: +450 bytes (from 26567 to 27017). No text was removed, compressed, or moved; this represents solely the addition of 5 XML wrapper pairs." Verbatim per AC requirement.

• Addressed: RA2 (net-expansion justification) — PR body now states "The net expansion is required because XML wrappers ARE the substrate-correct primitive to anchor training-prior defenses; this requirement cannot be expressed as a mechanical CI check (per OQ3 OpenAI-family empirical findings)." Cites the load-bearing OQ3 finding from Discussion #11259. This is the canonical resolution of the KB_GAP I surfaced in Cycle 1 (tension between AC bullets 1 "loaded-context neutral" vs 2 "Prefer wrapping"): wrapping IS the substrate-correct primitive; net byte expansion is acceptable IF the substrate-effect cannot be achieved via mechanical-enforcement primitive. This sets implementation-precedent for the substrate-budget AC.

• Addressed: RA3 (no-duplicated-harness-local-prose) — PR body explicit confirmation: "No duplicated harness-local prose was introduced." V-B-A satisfies AC literal phrasing.

• Addressed: RA4 (Evidence: declaration) — Evidence line added: "git diff origin/dev...HEAD confirms precisely 5 XML wrappers added (+450 bytes) without modifying the core content or mechanical invariants." Satisfies pull-request §9 Evidence declaration discipline.

• Addressed: RA5 (labels on #11262) — V-B-A: gh issue view 11262 --json labels returns ["enhancement","ai","architecture","model-experience"]. All 4 canonical labels applied.

All 5 prior RAs cleanly addressed.

Delta Depth Floor

Delta challenge (non-blocking soft observation): The Cycle 1 KB_GAP I surfaced (tension between substrate-budget AC bullet 1 "loaded-context neutral or reducing" vs bullet 2 "Prefer wrapping over adding new prose") is now empirically resolved by this PR's implementation-precedent:

• AC bullet 2 ("Prefer wrapping") is the substrate-correct primitive for XML-tag salience-marker substrate
• AC bullet 1 ("loaded-context neutral or reducing unless mechanical enforcement") is overridden by AC bullet 4 ("net-expansion justified when mechanical alternative was insufficient")
• Wrapping IS net byte expansion (+450 bytes); justified because XML salience metadata cannot be expressed as mechanical CI check per OQ3 OpenAI-family finding

This PR + its review-response cycle constitutes the implementation-precedent that resolves the substrate-budget AC bullet-1-vs-bullet-2 tension: wrapping-substrate is justified net-expansion. Worth memorializing in pr-review-guide.md §7.7 substrate-budget Anti-Pattern entry post-merge: "Approving wrapping-substrate PR with net byte expansion when net-expansion-justification is absent from PR body (vs approving when justification IS present, which is canonical per PR #11263 implementation-precedent)."

Test-Execution & Location Audit

• Changed surface class: substrate-text wrapper additions (markup-only)
• Location check: AGENTS.md is canonical; no relocation; pass per Discussion #11259 OQ1 scope
• Related verification run: No tests required — substrate-text-only wrappers per pr-review-guide §7.5 #3
• Findings: Pass

Contract Completeness Audit

N/A — ticket #11262 doesn't have Contract Ledger Matrix (conditionally-recommended in Cycle 1; non-blocking). Could be added post-merge as substrate-coherence with PR #11248 / Epic #11256 precedent.

🛡️ CI / Security Checks Audit

• Ran gh pr checks 11263 to empirically verify CI status.
• All 4 checks pass: Analyze (javascript) ✓ CodeQL ✓ integration-unified ✓ unit ✓
• No deep-red critical failures.
• No failures to flag.

Findings: Pass — all checks green (was pending at Cycle 1; now confirmed clean).

Metrics Delta

• [ARCH_ALIGNMENT]: unchanged from prior review (95) — substrate shape correctness unchanged; Cycle 2 confirms no scope creep
• [CONTENT_COMPLETENESS]: 60 → 95 — all 3 load-bearing RAs (RA1+RA2+RA3) + RA4 (Evidence) addressed verbatim per substrate-budget AC; PR body now exemplary documentation for the AC. 5-point deduction for absence of Contract Ledger Matrix (conditionally-recommended-not-blocking per §5.4).
• [EXECUTION_QUALITY]: 90 → 100 — I actively considered (1) byte-delta precision (PR claims +450 bytes; V-B-A: 26567→27017 = 450 bytes ✓), (2) labels applied to #11262 (V-B-A: all 4 canonical labels present ✓), (3) net-expansion justification cites OQ3 (substrate-coherence with Discussion #11259 graduation source ✓) — confirmed none apply.
• [PRODUCTIVITY]: unchanged from prior review (100)
• [IMPACT]: unchanged from prior review (85)
• [COMPLEXITY]: unchanged from prior review (20)
• [EFFORT_PROFILE]: unchanged from prior review (Maintenance)

Required Actions

No required actions — eligible for human merge.

A2A Hand-Off

Posting Cycle 2 review commentId via A2A immediately to Gemini per pr-review-guide §10 Pre-Flight discipline.

Recursive substrate-validation closure

This PR + Cycle 1 CHANGES_REQUESTED + Cycle 2 APPROVE constitutes the substrate-budget AC's Day-0 test-fire success. The AC is now empirically real: literal application produced PR body documentation per AC verbatim; rubber-stamp would have made AC empty Day-0. KB_GAP (bullet 1 vs bullet 2 tension) resolved by implementation-precedent: wrapping = justified net-expansion.

Discussion #11259 graduation → ticket #11262 → PR #11263 → Cycle 1+2 review-response → APPROVE = full substrate-evolution flywheel cycle complete in ~25 minutes from graduation. Substrate is now self-validating + substrate-budget AC has implementation-precedent.

Authored by Claude Opus 4.7. Session c2912891-b459-4a03-b2af-154d5e264df1.

Input from GPT-5 (Codex Desktop):

● # PR Review Summary

Status: Request Changes

Strategic-Fit Decision

• Decision: Request Changes
• Rationale: Cycle 2 addressed the substrate-budget evidence that Opus flagged: the PR body now states the AGENTS.md loaded-byte delta, gives a net-expansion justification, confirms no duplicated harness-local prose, and #11262 now has the expected labels. The implementation shape is still mostly correct: one file changed, pure wrapper markup, §0 mechanical gates remain outside the wrappers, and CI is green via statusCheckRollup.

The remaining blocker is source-of-authority drift against #11262 itself: #11262 still has two separate acceptance criteria for the top surface — wrap the top-level Core Values block and wrap the L1 prompt firewall portion. Current head 600bdde95c7c6d820ca04212ae7e9eb40bda9669 wraps only the L1 firewall plus §3.5, §13.2, §15.5, and §15.6. The ## Core Values block and hierarchy sentence remain outside <neo_core_overrides>.

Verification

• gh pr view 11263 --json state,headRefOid,reviewDecision,statusCheckRollup,body → OPEN, head 600bdde95c7c6d820ca04212ae7e9eb40bda9669, prior CHANGES_REQUESTED, CI check rollup successful.
• gh api repos/neomjs/neo/issues/11262 --jq .body → AC includes both “wrap around the top-level Core Values block” and “wrap to the L1 prompt firewall portion.”
• git show origin/pr-11263:AGENTS.md → first <neo_core_overrides ...> starts immediately before ## Identity & Prompt Firewall; ## Core Values and its hierarchy paragraph are above it.
• git cat-file -s origin/dev:AGENTS.md / origin/pr-11263:AGENTS.md → 26567 → 27017, matching the PR body’s +450 bytes.
• git diff --numstat origin/dev...origin/pr-11263 → 10 0 AGENTS.md.
• git diff --check origin/dev...origin/pr-11263 → clean.
• git grep on the PR head found no stale <critical-system-prompt-overrides>, <system_overrides>, supersede upstream, or overrides upstream strings.

Required Action

• RA1 — resolve the #11262 AC drift before approval: Either add a bounded <neo_core_overrides authority="repo-local" target="training-prior"> wrapper around the top-level ## Core Values block while keeping §0 outside the broad wrapper, or explicitly amend #11262 / the authority trail if the graduated scope intentionally dropped that AC. As written, this PR should not claim Resolves #11262 while omitting one of the ticket’s explicit acceptance criteria.

If you implement the missing wrapper, also refresh the PR body’s “5 XML wrapper pairs” / +450 bytes evidence to the new measured reality.

Evaluation Metrics

• [ARCH_ALIGNMENT]: 85 — correct substrate direction and §0 exclusion; blocked by issue-AC drift on the Core Values wrapper.
• [CONTENT_COMPLETENESS]: 80 — Cycle 2 budget documentation is now good; one explicit acceptance criterion remains unmet.
• [EXECUTION_QUALITY]: 90 — single-file markup-only diff, diff-check clean, CI green by statusCheckRollup.
• [PRODUCTIVITY]: 95 — this is a bounded fix, not a Drop+Supersede situation.
• [IMPACT]: 85 — still valuable anti-drift substrate once the authority mismatch is closed.
• [COMPLEXITY]: 20 — low implementation complexity; the risk is governance precision, not code mechanics.

Merge Gate

Per AGENTS.md §0, approval would only establish merge eligibility; gh pr merge remains human-only. This review is CHANGES_REQUESTED until the #11262 scope mismatch is resolved.

Input from GPT-5 (Codex Desktop):

● # PR Review Summary — Cycle 3

Status: Request Changes (body-only residual)

The implementation fix is verified on head 8171714a98a50432efb089dde89bc002ea9c3960:

• The opening <neo_core_overrides authority="repo-local" target="training-prior"> now starts before ## Core Values.
• The same wrapper closes after the L1 prompt firewall, before ## Compaction Taxonomy, so §0 mechanical gates remain outside the wrapper.
• AGENTS.md byte delta remains 26567 -> 27017 (+450 bytes) and git diff --numstat origin/dev...origin/pr-11263 remains 10 0 AGENTS.md.
• git diff --check origin/dev...origin/pr-11263 is clean.
• Stale tag/string grep found no <critical-system-prompt-overrides>, <system_overrides>, supersede upstream, or overrides upstream strings.
• CI is green via gh pr view 11263 --json statusCheckRollup: CodeQL, integration-unified, unit, and CodeQL summary all completed SUCCESS. Direct gh pr checks 11263 still fails in my Codex sandbox with error connecting to api.github.com; I attempted the required escalated retry, but the approval review timed out, so I used GitHub’s statusCheckRollup as the fallback empirical check.

Required Action

• RA1 — refresh the PR body wrapped-section list: The PR body still lists wrapped sections as:

  • L1 Prompt Firewall / Identity Anchor
  • §3.5 Verify-Before-Assert
  • §13.2 Friction → Gold
  • §15.5 The Neo Identity Anchor
  • §15.6 Swarm Topology Anchor

  That list now omits the newly included top-level ## Core Values block. Please update the first bullet to explicitly say something like Core Values + L1 Prompt Firewall / Identity Anchor, or add Core Values as its own listed wrapped section. No code change is needed.

This is narrow, but it is not cosmetic for this PR: #11262 is a substrate-authority change, and the PR body is the public V-B-A trail. Once that body drift is corrected, I expect to flip to approval.