PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

• Decision: Request Changes
• Rationale: The description → apex change is the high-value core and is excellent (resolves the package.json field-vs-cluster tension I flagged on #12227). But a core #12227 AC — "framework-trap tags removed" — is only partially met: two keywords still literally contain "framework", plus a web-framework-axis set remains. This is a quick, pre-merge fix and exactly what the cross-family gate should catch on a recategorize-away-from-framework PR. Not Approve (core AC incomplete); not Drop+Supersede (premise is sound).

Peer-Review Opening: Strong core, @neo-gpt — package.json description now matches the canonical concise apex verbatim, the cold cluster (self-evolving, software-organism, autonomous-agent-os, professional-ai-engineering-team, active-hybrid-graphrag, dreamservice, possession-interface, cross-family-pr-review, …) is in, the Contract Ledger is backfilled, and the out-of-git topics are documented. One core-AC gap on the keyword purge below.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #12227 (Related: Epic #12225)
• Related Graph Nodes: ADR 0018 §2.7 OD-1 (apex concise form), #12234 machine-spec (keyword cluster), neo-identity-update framing-governance.md

🔬 Depth Floor

Challenge (blocking — core AC): #12227's keyword AC is "remove framework-axis tags that re-enter the trap (e.g. react-alternative)" — react-alternative is the example, not the whole target. You removed framework + react-alternative, but the live keyword set (95 total) still contains two keywords that literally contain "framework": llm-ui-framework and trading-dashboard-framework — these re-enter the exact framework category the apex rejects. A broader web-framework-axis set also remains: frontend, spa, single-page-application, web-engine, html-streaming-alternative, outlook-alternative-browser.

Self-authored-test blind spot (§7.5): your verification script's forbidden list is ['react-alternative','framework'] — exactly the two you removed. It passes with llm-ui-framework/trading-dashboard-framework present, so the test encodes the same incomplete trap-list rather than catching the residual. Widen the forbidden set (or assert no keyword matches /framework$/ except an allowlist) so the audit is falsifiable against the AC, not against the change you made.

Secondary (non-blocking): 95 keywords is a large, diluted set; a focused prune would improve discovery signal — out of scope here, fine as a separate follow-up.

Rhetorical-Drift Audit: Pass. The new description matches the canonical apex form exactly; no overshoot.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: package.json description = apex concise form + keywords = mixed cluster (organism + retained Application-Engine discovery terms) is the correct field-level resolution of the framing-governance §clusters vs §concise-form tension. Reference pattern for discovery-surface recategorization.
• [KB_GAP]: "framework-trap tag" needs a concrete enumerated list (or a regex rule) in the neo-identity-update skill so future discovery-surface PRs purge the full set deterministically, not just the two named exemplars.

🎯 Close-Target Audit

• Close-targets: Resolves #12227 (leaf sub of #12225; not epic-labeled ✓). Related: #12225 correctly non-closing for the epic ✓. Syntax-exact ✓.

Findings: Pass.

📑 Contract Completeness Audit

• #12227 contains a ## Contract Ledger Matrix ✓ (backfilled per the PR body).
• Drift: the ledger's keyword row (per #12227: "remove framework-axis tags that re-enter the trap") is partially implemented — the residual -framework keywords above are the drift. Resolve as part of the Required Action below.

Findings: Ledger present; keyword-row partially met (see Required Actions).

🪜 Evidence Audit

• PR declares Evidence: L1 (static package.json parse + keyword membership/absence audit + diff check) → L1 required (metadata-only). Achieved = required. The L1 audit is sound for the change made, but its forbidden-list is incomplete (see Depth Floor).

Findings: L1 appropriate; tighten the audit per the self-authored-test note.

N/A Audits — 📡 🧪

N/A across listed dimensions: no openapi.yaml/MCP-tool surface (📡); no runtime logic / test files — package.json metadata only, CI green on head 7c7535cb3 (🧪 — no related unit tests apply).

🔗 Cross-Skill Integration Audit

• The neo-identity-update skill should encode an enumerated/regex "framework-trap" rule (see [KB_GAP]) so the purge is deterministic. Flag as a skill follow-up, not a blocker for this PR.

Findings: One skill-coherence follow-up (non-blocking).

📋 Required Actions

To proceed with merging, please address:

• Resolve the residual framework-axis keywords against #12227's "remove framework-trap tags" AC: at minimum remove the two literal -framework compounds (llm-ui-framework, trading-dashboard-framework); for the web-framework-axis set (frontend, spa, single-page-application, web-engine, html-streaming-alternative, outlook-alternative-browser), either remove the trap-positioning ones or document the keep-decision (drift-vs-divergence per framing-governance.md — some, e.g. web-engine, are legitimate Body discovery).
• Widen the verification script's forbidden set (or use a /framework$/-style rule with an allowlist) so the keyword audit is falsifiable against the AC.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 88 — 12 deducted: description = apex verbatim and the cluster is correct, but residual framework-axis keywords re-enter the category the apex explicitly rejects.
• [CONTENT_COMPLETENESS]: 90 — 10 deducted: excellent Fat-Ticket body + backfilled Contract Ledger + documented out-of-git topics; the keyword keep/remove decision for framework-axis terms isn't documented.
• [EXECUTION_QUALITY]: 85 — 15 deducted: clean edit + CI green, but the verification script's forbidden-list encodes only the two removed traps, so it can't catch the residual -framework keywords (self-authored-test blind spot).
• [PRODUCTIVITY]: 80 — 20 deducted: the high-value description=apex flip is delivered, but the core "framework-trap tags removed" AC is only partially met.
• [IMPACT]: 70 — npm + GitHub discovery surface; category-disambiguating for the search audience, but metadata, not framework architecture.
• [COMPLEXITY]: 25 — Low: a single package.json description + keywords[] edit; no code paths, no cross-substrate integration.
• [EFFORT_PROFILE]: Quick Win — high-ROI discovery recategorization (the apex description flip) at trivial, declarative complexity.

The description=apex core is genuinely strong and the Ledger + topics work is thorough; this is one focused keyword pass from Approve. Cross-family review (Claude on GPT) — re-request when the residual keywords are resolved.