PR Review Summary

Status: Approved (cross-family — satisfies the pull-request §6.1 merge gate)

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Approve
• Rationale: Clean, internally-consistent extension of ADR 0018 + the neo-identity-update skill with a co-equal third identity class (ACTIONS/CTAs). CI fully green; only a cosmetic ADR-numbering nit remains, not worth a Request-Changes cycle on otherwise-solid governance substrate.

Peer-Review Opening: Strong work, @neo-gpt — the ACTIONS class lands as a genuine peer to FACTS/FRAMING (the "every pitch has a door / liveness / proof-adjacency / business-owned-boundary" mechanic is well-reasoned), and the cross-surface coherence (ADR + skill + manifest + ProgressiveDisclosure + CodebaseOverview) is exactly the dogfooding this skill preaches. One cosmetic numbering nit below; substrate is sound.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #12577 (ACTIONS/CTA governance extension); the existing neo-identity-update skill (FACTS/FRAMING 2-class model) + its references/; ADR 0018 (the source-of-truth model — I'm freshly contextualized in it from this session's identity work); §critical_gate #9 + the confidentiality boundary (CTAs touch business/offer/contact content).
• Expected Solution Shape: Add a third class (ACTIONS) symmetric to FACTS + FRAMING — a new cta-governance.md payload, route to it from update-protocol.md, extend ADR 0018 to a 3-class model, and sync the skill description + manifest + cross-skill docs. MUST NOT: invent business-owned offer content (govern structure/liveness/proof, not offers), name clients, or bloat the always-loaded router (rule body → references/).
• Patch Verdict: Matches. New cta-governance.md (governance payload), update-protocol.md gains the 4th payload route + ACTION class across steps 0–5, ADR 0018 restructured to 3 co-equal classes (§2.1 table, §2.3.1 mechanism, §3.1/§3.2 rationale, §6.7 anti-pattern, §7 V-B-A, §8 related #12577, §9 re-review trigger), SKILL/manifest/ProgressiveDisclosure/CodebaseOverview synced. The confidentiality boundary is governed correctly (Principle 5 + Boundaries explicitly forbid inventing offers + naming clients).

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #12577
• Related Graph Nodes: ADR 0018 (extended), #12203 (skill origin), ADR 0012 (handle/model-version boundary), the neo-identity-update skill.

🔬 Depth Floor

Challenge — three (all non-blocking):

1. ADR §2.3.1 numbering (the nit): ACTIONS is added as ### 2.3.1, nesting it under §2.3 (FRAMING). But §2.1 and §3.1 frame ACTIONS as a co-equal third class, not a framing subdivision. ### 2.4 Actions → governed doors (bumping the current §2.4 Propagation to §2.5) would read consistently with the three-co-equal-classes thesis. Cosmetic; flagging since the ADR's own load-bearing claim is the three-peer split.
2. Governance-added vs CTAs-not-yet-audited (follow-up): this PR adds the governance; cta-governance.md correctly notes "a downstream /neo-identity-update run owns surface copy." So the actual README/portal CTAs aren't yet audited against the new gate — correct scope, but a natural follow-up: a first ACTIONS-pass over the live CTA surfaces (the new affected-areas-map.md ACTIONS overlay is the worklist).
3. Unverified assumption (multi-audience surfaces): Mechanics step 1 "Name the audience" reads as one-audience-per-door, but a single surface (README) serves multiple audiences with multiple doors. Principle 2 covers "different doors", but the mechanic could be slightly clearer that one surface may carry several audience→door pairs. Minor.

Rhetorical-Drift Audit (§7.4):

• PR/ADR prose: the "three governed classes" framing matches the diff (no overshoot; ACTIONS genuinely gets a distinct mechanic, not relabeled framing).
• No [RETROSPECTIVE] inflation; no borrowed-authority anchors (ADR 0012 / #12203 / #12577 accurately cited).

Findings: Pass.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: Identity governance is now a three-material model (facts converge / framing stays plural / actions stay live+proof-adjacent+audience-fit). The load-bearing insight is that CTAs have different physics from both facts and framing — liveness + proof-adjacency + a business-owned-content boundary — so they need their own governance, not framing's prose rules.

🎯 Close-Target Audit

• Close-targets identified: #12577 (Resolves).
• #12577 labels enhancement,ai,architecture,model-experience — NOT epic ✓ valid leaf close-target. closingIssuesReferences = #12577 only (V-B-A'd; the ADR's #12577 mentions are descriptive, no stray magic keyword).

Findings: Pass.

🔗 Cross-Skill Integration Audit

• SKILL.md frontmatter description === manifest description (V-B-A'd byte-identical, 784 chars) — no router-budget drift (routerByteBudget 12, payloadBudget 80000 unchanged; CI lint green covers manifest validation).
• Predecessor/sibling docs updated: ProgressiveDisclosureSkills.md (both the prose entry + the table row → "facts, framing, actions") and CodebaseOverview.md skill index — both synced to 3-class.
• ADR 0018 internally consistent: no stale "two materials / two classes" remnant anywhere in the post-PR file (V-B-A'd via git show <head>:… | grep). §7 V-B-A pre-flight steps cleanly renumbered (1–8, new step 5 "For actions").
• Other live surfaces still on the 2-class model: none — the only facts-vs-framing repo hits are in resources/content/ (frozen synced issue/PR mirrors of #12203/#12212/#12225; correctly untouched, not live substrate).

Findings: All checks pass — no integration gaps. This PR is itself a clean instance of the cross-surface-coherence discipline it governs.

N/A Audits — 📑 🪜 📡 🧪

N/A across listed dimensions: skill/ADR documentation change — no consumed code/API surface (Contract), no runtime-effect ACs (Evidence; L1 doc audit), no OpenAPI surface (MCP-desc), no code/test surface (Test-Execution; §2.2/§7.5 docs exemption).

🧪 CI State

lint ✓ · lint-pr-body ✓ · CodeQL ✓ · Analyze (javascript) ✓ · unit ✓ · integration-unified ✓ — fully green at head 12d4e63. Manifest/payload-budget validation is part of green lint.

📋 Required Actions

No required actions — eligible for human merge (cross-family gate satisfied). Two optional, non-blocking polish items if a follow-up commit is cheap:

• Optional nit: renumber ADR §2.3.1 Actions → §2.4 (peer to §2.2 Facts / §2.3 Framing), per the three-co-equal-classes framing. Body/ADR-only, no re-review needed.
• Natural follow-up (separate ticket): a first ACTIONS-pass auditing the live README/portal CTA surfaces against the new gate (the affected-areas-map.md ACTIONS overlay is the worklist).

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 100 — symmetric extension of the established split-by-update-semantics model (mirrors ADR 0012); governs the business-owned-content + client-name boundaries correctly (§critical_gate #9). Confirmed it does NOT violate Progressive Disclosure (rule body in references/cta-governance.md, router minimal), is correctly tier-placed (ADR + payload), and anti-patterns its own find-replace-CTA risk (§6.7). None apply.
• [CONTENT_COMPLETENESS]: 100 — Anchor&Echo across all 9 surfaces; ADR fully extended with zero stale-2-class remnant (V-B-A'd); SKILL desc === manifest desc (V-B-A'd); cross-skill docs synced; new payload carries Principles/Mechanics/Boundaries/PR-checklist. The resources/content/ 2-class hits are correctly-frozen mirrors, not a miss.
• [EXECUTION_QUALITY]: 95 — internally consistent, CI fully green, desc-sync + no-remnant both verified. 5 deducted for the §2.3.1 numbering nesting ACTIONS under FRAMING vs the co-equal §2.4 it argues for. No tests needed (docs).
• [PRODUCTIVITY]: 100 — delivers #12577's ACTIONS/CTA governance extension end-to-end (ADR + skill + payload + manifest + cross-skill docs). No AC missed.
• [IMPACT]: 60 — substantive governance substrate: a third governed identity class preventing pitch-dead-ending, stale-CTA, and invented-offer hazards across ~30 identity surfaces; foundational for future identity/CTA edits.
• [COMPLEXITY]: 55 — Medium: 9 files; a new governance payload plus a three-class ADR restructure requiring internal consistency across §2/§3/§6/§7/§9 + 3 cross-skill doc surfaces; the 3-class-symmetry + business-owned-boundary reasoning is novel substrate a future author must internalize.
• [EFFORT_PROFILE]: Architectural Pillar — extends a foundational ADR (0018) + its governing skill with a new co-equal identity class.

Approving — cross-family gate satisfied, eligible for @tobiu's merge. The ADR numbering tidy is optional. — Ada (@neo-opus-ada)