Asset template audit: anchor-preserving first-pass vs follow-up split

Context

Sub 5 of Epic #10733. Depends on Sub 1 baseline. Templates are the most easily-extracted-but-most-easily-broken surface — their structural anchors are load-bearing for the Retrospective daemon's regex parser.

The Problem

Asset templates (line counts from local sweep):

pr-review-template.md (216 lines) loads cold-cache on every Cycle 1 PR review. Sections within it (e.g., Source-of-Authority audit, Provenance Audit) may be condition-gated rather than universal. But — per @neo-gpt's Depth Challenge — templates are graph-ingestion + review-normalization surfaces. Section anchors and labels are load-bearing for the Retrospective daemon's regex parser. Splitting without anchor preservation breaks graph ingestion silently.

The Architectural Reality

• .agents/skills/pr-review/assets/pr-review-template.md — Cycle 1 cold-cache template
• .agents/skills/pr-review/assets/pr-review-followup-template.md — Cycle N warm-cache template (split already exists at the template-shape level; this sub asks whether the section-level split inside pr-review-template.md is also justified)
• .agents/skills/epic-review/assets/epic-review-comment-template.md — epic-review template
• Retrospective daemon regex parser — ingests structured sections by Markdown header; load-bearing for graph-ingestion (per feedback_pr_review_template_discipline)

The Fix

Conduct a parser/anchor audit for each candidate template before any split. Enumerate which structural anchors the Retrospective daemon's regex parser depends on. Templates are not just prose; they are graph-ingestion contracts.

First-pass vs follow-up section-level split is the most obvious candidate for pr-review-template.md — subsequent reviews rarely need the full provenance audit, and the cold-cache cost on Cycle 1 dominates the cumulative loaded-byte count. But the split must preserve stable anchors.

Lazy-load verification gate (per @neo-gpt): the per-harness loaded-byte delta of any proposed split must be empirically positive (not just smaller files on disk).

Acceptance Criteria

• (AC15) Parser/anchor audit completed for each candidate template; anchors required by the Retrospective daemon enumerated
• (AC16) First-pass vs follow-up split executed for pr-review-template.md if the audit supports it; structural anchors preserved across the split; per-harness loaded-byte delta verified positive
• (AC17, [L2-L3-deferred — operator handoff needed]) Post-split: Retrospective daemon ingestion verified by sample-ingest of a recent PR review. AC17 will remain unchecked at merge by design — verification requires actually running IssueIngestor on a recent PR review (e.g. PR #10752 Cycle 1 review at PRR_kwDODSospM78GE72); this is L2-L3 evidence per evidence-ladder.md and is sandbox-unreachable from agent-side PR review. Tracked in PR #10754 Post-Merge Validation per pull-request-workflow §9 standard pattern.

Out of Scope

• AGENTS.md, boot ramp, or skill-payload changes (Subs 2-4 own)
• Replacing template prose with structured data (YAML, JSON, XML) — out per Epic Out-of-Scope
• Authoring new template anchors not currently parsed — anchors enumerated in AC15 are the floor, not a target for expansion

Related

• Parent Epic: #10733
• Origin discussion: #10732 (especially @neo-gpt's templates-as-graph-ingestion-surface framing)
• Architectural reference: Retrospective daemon graph parser (see feedback_pr_review_template_discipline)

Origin Session ID: 7e52099b-9632-4c67-a2a1-4e1a1ad1c414

Retrieval Hint: query_raw_memories(query="asset template audit pr-review-template anchor-preserving first-pass follow-up split graph-ingestion 10733 10732")