Codex Phase 3 Response — STANDARDS.md § 11 Enforceability
Date: 2026-05-14 Model: gpt-5.5 (via mcp__codex__codex) Prompt sent: Phase 3 enforceability prompt + the new STANDARDS.md § 11 section Thread: 019e274e-b17a-7f43-8e7b-da2f32e374fd
1. Enforceable Items
R4, A1, A6, A7, S3, C3, C5, C6
2. Ambiguous Items + Suggested Rewording
R1 — "topic + stakes" and "no jargon" are subjective — "First 80 words must name the foundation topic, state one concrete consequence of poor preparation, and contain no undefined terms from project glossary or domain-specific acronyms."
R2 — "reachable in mobile first viewport" depends on viewport/device — "At 390x844 viewport, at least one link or CTA labeled 'Start here' / 'What to do next' must be visible without scrolling."
R3 — "reader questions" vs "topic taxonomies" is vibes — "Every H2 must be phrased as a question or action-oriented reader task, not a noun-only category label."
R5 — "before abstract framing" is unclear — "At least one concrete numeric quantity must appear within the first 150 body words, excluding frontmatter."
R6 — "insider term" undefined — "Any term listed in glossary/domain vocabulary must be followed within the same sentence by a plain-language definition."
A2 — "major claim cluster" and "weighted" are undefined — "Each H2 section containing factual, safety, legal, medical, technical, or procurement claims must include at least one citation meeting SOURCES.md minimum tier for that claim type."
A3 — breadth options are unclear — "Hub must include either all three setting labels Urban, Suburban, Rural, or all three duration labels Emergency, Short-term, Permanent, with at least one actionable difference under each."
A5 — "depth signal" is not testable — "Each major H2 must include at least two of: numeric threshold, decision rule, named method, material/tool specification, comparison table, or step sequence."
A8 — "expert help threshold" vague — "Life-safety hubs must include an H2 or callout titled When to get expert help with at least three concrete trigger conditions."
S1 — "answers query intent" subjective — "Meta description must be ≤160 characters and state the page's primary task/query in plain language."
S2 — "passage-ranking lens" is meaningless operationally — "Each H2 must cover one distinct subtopic; H3s must answer specific sub-questions under that H2. No H2 may contain more than 400 words without an H3."
S4 — "schema-eligible" and "schema stuffing" need exact schemas — "Must include one visible FAQ, HowTo, BreadcrumbList, or ItemList element and matching JSON-LD using the same visible text."
S5 — "featured-snippet candidate" vague — "Within first 250 body words, include one standalone paragraph of 40–50 words directly answering the hub's primary query."
S6 — "dominant search intent" external and unstable — "Editor must classify intent as one of informational, how-to, or comparison; page must contain matching structure: definition block, steps/checklist, or comparison table."
S7 — "adjacent query variants" undefined — "Include at least three exact subquery phrases from approved keyword list or editor-provided scorecard."
C1 — "above the fold" undefined — "At 390x844 viewport or within first two H2 sections, include 1–3 links labeled as start actions."
C2 — "audience-segmented" vague — "Include at least two labeled audience paths from an approved set, e.g. Beginner, Family, Apartment, Rural, Limited budget, Advanced."
C4 — "single next action" can be interpreted loosely — "Final section must end with exactly one internal link or checklist action, and no recap paragraph after it."
C7 — "beginner-to-depth progression visible" subjective — "Include an ordered path of at least three internal links labeled beginner, intermediate, and advanced."
Editor checklist tone item — "calm/practical/empowering" is subjective — "Ban specified phrases/categories: panic language, partisan labels, apocalyptic claims, moral judgment. Require neutral instructional wording."
3. Missing Enforcement Detail
Hard-fail scope — R4 is hard-fail in list but not labeled [HARD-FAIL] in rubric — add [HARD-FAIL] inline to R4.
A4 — missing definition of "life-safety claims," "weighted," and "warned" — add claim categories, required source tiers, and required warning/callout format.
A6 — missing classification of high-criticality vs standard hubs — add table mapping each hub to 90-day or 180-day requirement.
A8 — "waivable elsewhere" conflicts with "every applicable item" — define who can waive, where waiver is recorded, and allowed waiver reasons.
S1 — hard-fail says "description ≤160 chars + H1 present," but rubric also says title match and query intent — publisher hard-fail must state exact checks.
S3 vs A1 — duplicate internal-link requirements conflict for hubs with fewer than 8 child pages — define exception or require all child pages plus enough cross-links.
S4 — JSON-LD validation method missing — require a specific script/check or schema validator and exact failure conditions.
S7 — no source of query variants — add approved keyword source or require variants to be listed in scorecard before review.
"Major claim cluster" — no unit of review — define whether clusters are H2 sections, paragraphs, tables, or callouts.
"Named institution" — undefined — specify acceptable institutions: government agencies, universities, standards bodies, hospitals, NGOs, manufacturers? Exclude blogs/forums?
"Mobile first viewport" / "above the fold" — no viewport — standardize viewport size and whether nav/header counts.
Waivers — soft-fail waiver process too loose — require rubric ID, failed evidence, reason, owner, and follow-up issue/backlog link.
Publisher enforcement — unclear which checks are automated vs human — split publisher-blockable automated checks from editor-attested checks.
Retrofit protocol — sprint-specific and dated; not enforceability standard — move to project plan or add expiry/removal condition.