Codify Collapsed-N/A audit conciseness rule in pr-review templates

Context

Friction-to-gold surfaced from PR #11544 review 2026-05-17 ~18:56Z: I (Claude Opus 4.7) used a novel collapsed-emoji-header shortening (### 🛂 / 📑 / 🪜 / 📜 / 📡 / 🔌 / 🛡️ → "N/A or Pass across all audit dimensions") for the 7 audit dimensions that all evaluated N/A for a test-reliability fix. Operator @tobiu caught it 2026-05-17 ~19:18Z as a "very useful hallucination" — substrate-correct emergent shape, just not codified in template.

The Problem

.agents/skills/pr-review/assets/pr-review-template.md has 9 separate audit sections (lines 64, 87, 98, 120, 137, 153, 165, 179, 192). For PR types where 5-8 of those dimensions are N/A (test-reliability fixes, docs-only changes, micro-refactors), the cycle-1 template structure forces 5-8 separate N/A sections — inflates review byte-cost without substrate value.

.agents/skills/pr-review/assets/pr-review-followup-template.md has 3 audit sections (Test-Execution + Contract Completeness + 🛡️ CI/Security); narrower but still has 2-section-collapse opportunity.

Without codified shortening:

• Each reviewer might invent ad-hoc shortenings (cross-peer inconsistency)
• Reviewers default to full N/A-per-section bloat — drain-mode-velocity-negative, especially costly in long PR conversations (today's 11-cycle #11534)
• Graph-ingestion regex anchoring (V-B-A: currently grep-zero on audit-section titles in ai/daemons/services/ConceptDiscoveryService.mjs + ai/scripts/) is unaffected; future graph-evolution that DOES key off audit anchors would be incompatible with ad-hoc shortenings

The Architectural Reality

V-B-A on graph-ingestion impact (per operator-direction "we need to VBA if graph ingestion needs to get enhanced too. maybe not."):

• grep -rn "Provenance Audit|Contract Completeness|Evidence Audit|Source-of-Authority|MCP-Tool-Description|Wire-Format Compatibility|Cross-Skill Integration|Test-Execution|CI / Security" ai/daemons/services/ConceptDiscoveryService.mjs ai/scripts/ → zero matches
• The Retrospective daemon's regex anchors on the 7 evaluation-metric tags ([ARCH_ALIGNMENT], [CONTENT_COMPLETENESS], etc.) per manage_pr_review openapi description + #11491 mechanical body validation, NOT audit-section emojis
• Conclusion: codifying a collapsed-N/A pattern is graph-safe. No graph-layer change needed.

Template byte impact:

• pr-review-template.md current: ~14KB (Atlas-tier, conditionally-loaded; not byte-budget-constrained)
• pr-review-followup-template.md current: ~3.5KB
• Conciseness Rule addition: ~19 lines (cycle-1) + ~15 lines (follow-up) — Atlas-tier, low cost

The Fix

Add a "🧱 Conciseness Rule — Collapsed-N/A Audits" subsection to BOTH templates after the Graph Ingestion Notes / Depth Floor section, codifying:

Cycle-1 template (pr-review-template.md):

<h3 class="neo-h3" data-record-id="6">🧱 Conciseness Rule — Collapsed-N/A Audits</h3>

When 2+ of the audit dimensions below (🛂 Provenance, 📑 Contract Completeness, 🪜 Evidence, 📜 Source-of-Authority, 📡 MCP-Tool-Description Budget, 🔌 Wire-Format Compatibility, 🔗 Cross-Skill Integration, 🧪 Test-Execution & Location, 🛡️ CI/Security) all evaluate **N/A** for the PR scope, collapse them under a single section using the explicit canonical-anchor format:

`### N/A Audits — 🛂 📑 🪜 📜 📡 🔌 🔗 🧪 🛡️`
`N/A across listed dimensions: <one-line reason for the PR-scope justification>.`

(Substitute only the emoji subset that's actually N/A. Use spaces between emojis, single-space-emdash separator after `Audits`.)

**Mixing rule (CRITICAL)**: substantive audit sections — any non-N/A finding — MUST be expanded individually under their canonical header. Never collapse a dimension with actual findings.

**Graph-ingestion safety**: V-B-A confirmed.

**Empirical anchor**: PR #11544 review 2026-05-17 (Claude Opus 4.7) introduced organically. Operator-codified 2026-05-17 ~19:25Z.

Follow-up template (pr-review-followup-template.md): equivalent rule, narrower scope (3 audit sections; collapse when 2+ are N/A).

Acceptance Criteria

• pr-review-template.md gains a "🧱 Conciseness Rule — Collapsed-N/A Audits" subsection with canonical-anchor format
• pr-review-followup-template.md gains equivalent (narrower scope) Conciseness Rule
• Canonical anchor format documented: ### N/A Audits — <emoji-subset> (spaces between emojis, single-space-emdash separator)
• Mixing-rule documented: substantive findings ALWAYS expand individually
• Empirical-anchor cross-reference: PR #11544 review + operator codification direction

Out of Scope

• Graph-ingestion regex changes: V-B-A confirms not needed. Post-v13 follow-up if graph evolves to ingest audit-dimension-coverage.
• Retroactive review-rewrites: prior reviews stay as-is. Rule applies going forward.
• Beyond audit sections: other template structure (Strategic-Fit, Depth Floor, Required Actions, Evaluation Metrics) is out-of-scope.

Avoided Traps

• ❌ [enhancement] prefix in title: title carries subject; labels handle category
• ❌ Substrate-evolution scope creep: keep change strictly to the 2 template files
• ❌ Retroactive-encoding anti-pattern (per #10082/#11540 lesson 2026-05-17 ~19:38Z): this is FORWARD-PREVENTIVE substrate-discipline, NOT retroactive-encoding — the rule applies to all FUTURE reviews + every reviewer benefits. Distinct from the tech-debt-radar-bloat anti-pattern.

Related

• Empirical anchor: PR #11544 review (pullrequestreview-4306145233) — organic pattern emergence
• Operator codification direction: 2026-05-17 ~19:25Z + 19:28Z chat thread
• Drain-mode framing: long PR conversations (today's 11-cycle #11534) benefit disproportionately from cycle-1-review byte-cost reduction
• §13.1 contributions-over-commits: this is a substrate-discipline tooling improvement that compounds across future reviews — drain-mode-velocity-positive long-term

Origin Session ID: 39eee906-3fd4-424f-9348-828b46ece38c

Retrieval Hint: "Collapsed-N/A audit conciseness rule pr-review template"