# Changelog All notable changes to this document are recorded here. Format: `[VERSION]—YYYY-MM—Description` ## [1.9.18]—2026-06-08 - **TH B2C punctuation ban gains a proper-noun/title carve-out.** A banned mark (em-dash, en-dash, colon, semicolon) that is part of a proper noun, publication name, or work/study title (e.g., the colon in *CA: A Cancer Journal for Clinicians*) is intrinsic to the name, not authorial punctuation, and is preserved verbatim; where a destination platform filter cannot keep it, comma-substitution is the defined fallback, never strip-with-no-replacement. Closes the edge case engineering surfaced testing PGS-385 (the model was left stripping the mark or guessing a substitute). `docs/trend-hunter-b2c.md` §Punctuation gains an Exception paragraph; `_data/rules/conflict-register.yaml` `punctuation.banned_marks` TH-B2C override rationale carries the carve-out (register 1.0.3 → 1.0.4, injected into the writer prompt); `api/reference.json` gains `trend_hunter_b2c_banned_punctuation_exception`. Vendored into write-assist (rules.js). Version stamps 1.9.17 → 1.9.18. ## [1.9.17]—2026-06-05 - **L&E placement-test freeze captured in the data-universe-labeling governance (Amplitude L&E caveats).** Operational rule that follows from the broken L&E Amplitude integration: no L&E content-placement tests until the L&E data pipe is repaired. An L&E placement test currently can't be measured (integration broken pre-~Aug-2025; the dedicated `AMPLITUDE_LIFESTYLE_AND_ENTERTAINMENT_EVENTS_PROD` view stale since 2026-02-24), so its result would be uninterpretable; exclude L&E from the placement-test matrix until the pipe is fixed. Version stamps 1.9.16 → 1.9.17; `assets/sources/` re-propagated via build-downloads.sh. ## [1.9.16]—2026-06-02 - **TH B2C headline advisories added (flag-only) to the machine register.** Three new `_data/rules/conflict-register.yaml` rules (register 1.0.2 → 1.0.3), default off, `publication=trend-hunter-b2c` override on: `headline.intent_advisory` (labels What/Why/How + what it drives), `headline.name_expert_advisory` (body cites a credentialed source the headline omits → suggest surfacing), `headline.etk_advisory` (flags "Everything to Know" no-lift framing). All advisory/flag-only—never gating, never auto-rewrite. Vendored into write-assist (Layered Style Guidance runners) + mirrored in the data-headlines variant tool. Human-readable conflict-register.md + assets updated. Version stamps 1.9.15 → 1.9.16. ## [1.9.15]—2026-06-02 - **Trend Hunter B2C headline standards added (§10.4) from the TH Headline Analysis (368 articles, 2026-06-02).** As the official TH content team, the content strategist's report learnings are encoded as TH B2C platform-layer standards (platform supersedes persona/format/general; AP-compatible is the ultimate fallback where TH B2C is silent). **`docs/trend-hunter-b2c.md` §10.4 Headlines** gains five standards: (1) **length 90–104 character sweet spot** (614 avg PV; <75 drops to 352, 105+ to 340; corpus median 86, so flag sub-80), overriding the universal 80–100, with destination distribution-platform ranges (Apple News 90–120, SmartNews 70–90) taking precedence when platform-bound; (2) **punctuation ban** on em-dash (—), en-dash (–), colon (:), semicolon (;) in public-facing headlines + body—flag for manual removal, hyphens allowed, restructure rather than substitute; (3) **intent match** (What/Why = reach, How = depth); (4) **name-the-expert** (single strongest lever—name the credential in the headline); (5) **"Everything to Know" discouraged** as a fallback (no lift). Openings gains a **first-scroll-delivery** bullet (Experiences + travel round-ups land clicks but lose readers when the open doesn't pay off). **Machine-readable register** (`_data/rules/conflict-register.yaml`, register_version 1.0.1 → 1.0.2): `headline.char_count` gains a `{publication: trend-hunter-b2c}` override (90-104, data-validated); new `punctuation.banned_marks` rule (default empty—no universal ban; TH B2C override bans the four marks; scoped so Woman's World no-space em-dash + US Weekly AP-spaced em-dash keep their house styles). **`api/reference.json`**: `platform_char_counts.trend_hunter_b2c` (90-104) + `trend_hunter_b2c_banned_punctuation` list added. Scoped TH-B2C-only per Pierce 2026-06-02—other publications comply with their own standards; avoid the banned marks unless a governing standard requires them. Downstream tools (headline grader, variant tool, Layered Style Guidance) align to this standard. Version stamps bumped 1.9.14 → 1.9.15. ## [1.9.14]—2026-05-28 - **§3.1 Google Discover Explainer retired (superseded by §3.12 What to Know Next).** Format moved to bottom of Article Formats list across all surfaces; marked `(retired)` in the same visual convention as `(pending)` formats. **Section anchors preserved** — `master-reference.md#31-google-discover-explainer` and `/docs/discover-explainer` URLs continue to resolve to the (now-retired) spec; historical inbound links remain valid. **Format-page status banner** added to `docs/discover-explainer.md` H1 + at top of body: "Status: Retired (2026-05-28). No longer in active production. Existing articles created under this spec remain valid; new content must use an active format. For forward-looking explainer needs, see §3.12 What to Know Next." **Navigation:** `_data/navigation.yml` entry moved from position 3 (alphabetic active group) to position 12 (after all pending formats); label updated to `(retired) Google Discover Explainer` mirroring the `(pending) X` convention. **API:** `api/reference.json` discover-explainer entry moved to bottom of `formats` array; `status: "active"` → `status: "retired"`; new fields `retired_on: "2026-05-28"` + `supersedes_by: "what-to-know-next"` added. **Documentation:** `docs/documentation.md` format count "4 active, 8 pending" → "3 active, 8 pending, 1 retired"; format table reordered (active first, then pending, then retired) with explicit note that section anchors reflect introduction order, visual ordering reflects status; `index.md` mirror count update. **Cross-references updated**: 7 docs surfaces touched — `docs/everything-to-know.md` (3 do-not-confuse refs), `docs/what-to-know-next.md` (2 do-not-confuse refs), `docs/master-reference.md` (§3.1 heading + retired banner + §3.2 do-not-confuse ref + version stamps), `docs/layered-enforcement.md` (Article Format layer format list re-grouped active / pending / retired), `docs/discover-browser.md` (persona page note that historically paired with §3.1, paired with active formats going forward), `_includes/headline-formulas/discover-explainer.html` (summary line + retired banner), `_includes/headline-formulas/everything-to-know.html` + `_includes/headline-formulas/what-to-know-next.html` (do-not-confuse refs noted retired). **Machine-readable register:** `_data/rules/conflict-register.yaml` discover-explainer overrides (3 found: headline.formula + body.word_count + structure.required_article_structure) annotated with new `format_status: retired` + `retired_on: "2026-05-28"` + rationale fields; rules preserved intact for historical-article validation. Human-readable `docs/conflict-register.md` mirror updated with inline retired markers. `_data/rules/schema.yaml` format-list description re-grouped (active / pending / retired); format enum re-ordered (active first; `discover-explainer` last with comment noting retention for historical validation). Version stamps bumped 1.9.13 → 1.9.14 across master-reference + api/reference.json. `scripts/build-downloads.sh` re-ran to propagate from docs/ → assets/sources/. Driver: Pierce 2026-05-28 decision; `What to Know Next` (v1.9.13 final-draft state) now covers the forward-looking explainer use case Discover Explainer was previously authoritative for. ## [1.9.13]—2026-05-28 - **§3.12 What to Know Next format reshaped to final-draft state per content team lead's 2026-05-28 spec.** Three net-new REQUIRED structural elements added to the format: (1) **Status Bar** — italicized one-liner directly beneath the H1 carrying `Stage · Momentum · Forecast confidence`, with enumerated allowed values (Stage: Emerging/Building/Peaking/Cresting; Momentum + Forecast confidence: Low/Moderate/High); positioned as a forcing function (if a writer can't fill it in confidently, the topic is wrong for this format). (2) **Forecaster Pull Quote** — one named-forecaster quote per article, named/titled/affiliated, forward-looking (prediction, interpretation, directional read), NOT definitional or descriptive; pull-quote visual treatment, not inline attribution; placed in Section 2 or 3 wherever it fits the arc. (3) **Three Signals to Watch** — boxed callout closing Section 3 listing exactly three specific, monitorable signals (date, release, product launch, data point, observable change); vague signals ("watch the market," "keep an eye on inflation") disqualify. **Word count reshape:** target widened 600–900 → **700–1,000 words** AND proportionality switched from even (200–300 per section) → **inverted distribution** (≈25% Section 1 / 30% Section 2 / 45% Section 3). Per-section word-count guidance dropped entirely (confirmed by content team lead same-day after initial ship — percentages carry the proportionality rule, prescribing per-section word counts over-constrains the writer). Section 3 is now the longest section by design—that's the format's defining point. New Formatting Rules: "Invert the word count" (Section 3 must be longest; if Section 1 is longest, the piece is structurally an Everything to Know with a forecast bolted on) and "Use the Forecaster Pull Quote for direction, not definition." Tone language softened slightly ("explanatory without being academic" → "explanatory without being overly academic"). Article Structure code block expanded to call out Status Bar position, Pull Quote placement option, and Three Signals callout. What to Avoid table extended with 7 new failure modes covering all three new REQUIRED elements + the inverted-proportionality requirement. Pre-Publish Checklist updated with 5 new gates (Status Bar all three fields; inverted proportionality; Section 3 is longest; Forecaster Pull Quote attribution + forward-looking nature + visual treatment; Three Signals exactly-three + concrete + monitorable) and updated word-count line. Master-reference §3.12 condensed summary mirrored. Override-banner expanded to list the three new format-specific REQUIRED elements. Driver: content team lead authored final-draft spec; exec/leadership endorsed integration into CSA; Pierce-side rewrite handles CSA-prompt-deliverability framing (REQUIRED sections with general-style agent-audience tags, enumerated allowed values, qualifying/disqualifying examples, table-of-attributes form for enforceability). Version stamps bumped 1.9.12 → 1.9.13 across master-reference + api/reference.json. `scripts/build-downloads.sh` re-ran to propagate from docs/ → assets/sources/. ## [1.9.12]—2026-05-27 - **AP Stylebook 58th edition (2026) drift fix + AI vocabulary block + Native Americans/Indigenous framing refresh.** Driven by AP Stylebook 58th edition release email forwarded from TH product owner same day. (1) **Drift fix:** `docs/ap-compatible-thorough.md` (Words Not to Hyphenate + Informal references sections) — `health care`/`health-care` replaced with `healthcare` as closed compound in all uses, per AP 58th edition (this change actually landed in AP's April 2026 batch and had been stale in the corpus for ~2 months). Single-file scan confirmed no other drift hits across the 4-file AP-Compatible tier set. (2) **AI vocabulary block** added to `docs/ap-compatible-thorough.md` (`### AI and Computing Terms` subsection under Part 14 Scientific Terms, after Technology Terms) and `docs/ap-compatible-condensed.md` (`### AI and Computing Terms (per AP 58th, 2026)` after Brand Names and Trademarks). 7 entries: `AI agent`, `AI chips`, `AI model`, `AI slop`, `compute` (noun), `data center`, `vibe coding`. Encoded as derived McClatchy style rules (not verbatim AP text — AP body content cannot be ingested verbatim per exec/leadership 2026-05-27 directive). Key McClatchy-specific override: `AI slop` carries an explicit `do not use to describe McClatchy's own CSA-generated output` clause (semantic conflict; CSA output is editorially supervised, not slop). `compute` (noun) prefers `computing power`/`processing power` outside direct quotes per AP guidance to avoid verb-noun confusion. `vibe coding` prefers `using an AI coding tool` outside direct quotes. Block NOT added to `docs/ap-compatible-quick.md` (at-the-elbow tier, intentionally minimal). (3) **Native Americans/Indigenous framing refresh:** section title in `docs/ap-compatible-thorough.md` renamed `Native American / American Indian` → `Native Americans, Indigenous peoples` (matches AP 58th title); body expanded with sovereignty + self-determination framing emphasizing tribal-nation citizenship, treaty rights, jurisdiction, federal recognition. `docs/ap-compatible-condensed.md` Indigenous + Native American/American Indian bullets reworded to match. Existing usage guidance preserved (all three terms acceptable; defer to individual + community + tribal-nation preference; identify by tribal affiliation when possible — added Diné, Haudenosaunee to the example list). Master-reference Us Weekly `American Indian → Native American` substitution rule left intact as publication-specific style choice (out of AP 58th scope). Version stamps bumped 1.9.11 → 1.9.12 across master-reference + api/reference.json. `scripts/build-downloads.sh` re-ran to propagate from docs/ → assets/sources/; 48 files regenerated (most are em-dash-sweep sync corrections that were already pending). Driver: TH product owner email forward of AP Stylebook 58th edition release notification; ~30-min Style Checker audit identified actionable deltas. ## [1.9.11]—2026-05-14 - **Two new editorial-governance docs** + one conflict-register entry. (1) `docs/ai-content-vetting.md`: partner-feed AI-content vetting policy. 5-gate test (named credentialed author · substance floor · plagiarism + AI-content detection · editorial fit · cadence audit), quarantine path (24h content-team-lead review), hard-block path (multi-gate failure → exec/leadership feed-level review), per-partner stance table (Reuters trusted; Stacker / Field Level Media / Minute Media active vetting; Campus Insights TBD). Operational mechanism uses Sonnet-class scaled-content-audit prompt for G2+G4 LLM-screening; new `partner_content_audit` D1 table for ingest event logs. Retires ops-hub p15-partner "Draft AI vetting policy proposal" nextAction + clears the p15 "AI vetting policy not drafted" blocker. (2) `docs/bridge-content-rules.md`: bridge content (purple-tier USW/WW articles pointing back to TH deepener via `rel="canonical"`) governance. Codifies "stay within trend but exact-keyword match NOT required"—pairs with conflict-register entry override on `headline.h2_keywords` (`platform_content_type = bridge` → `trend-relevance-only`). Driven by 2026-05-14 bridge-similarity probe (32 bridge clusters mean head_blend 0.370 vs 33 O&O-only baseline 0.376—bridges differentiate at parity with O&O at the headline level; strict keyword rule was over-constraining without scaled-content benefit). Retires ops-hub p4-governance "Loosen CSA strict-keyword-rule for bridge variants" nextAction. Version stamps bumped 1.9.10 → 1.9.11 across master-reference + api/reference.json. Probe deliverable cross-linked from `data-headlines/dev-docs/bridge-variant-similarity-2026-05-14.md`. ## [1.9.10]—2026-05-13 - **Conflict-register accuracy pass** (driven by `write-assist` 10-agent + code-quality audits). Five register fixes + nine doc-prose fixes across `docs/ap-compatible.md`, `docs/conflict-register.md`, `docs/layered-enforcement.md`, `docs/master-reference.md`, `docs/us-weekly.md`, `docs/womans-world.md`, and this changelog. **Register-level**: (1) `punctuation.em_dash_surrounding_spaces` default inverted from `false` → `true` (the canonical `ap-compatible-quick.md` + `ap-compatible-condensed.md` Em Dash sections both say "Use spaces on both sides" and flag the no-space form as a "common error"; the register's prior `false` default contradicted its own cited source). USW override marked `kind: restate` since USW matches the corrected AP default; **new WW override added** with `value: false` (WW's actual house style is no-space em dashes, an AMI Women's Group house-style **deviation** from AP, not a match as the WW doc had previously claimed). (2) `punctuation.ellipsis_surrounding_spaces` default same-direction inversion (`false` → `true`); USW override marked `kind: restate`. (3) `voice.banned_phrasing_demographic`—added `persona: curious-optimizer` override mirroring the TH B2C banned-phrase list, so the rule fires for persona-only selectors (the persona is psychographic by design and demographic stereotyping contradicts it). (4) Four redundant overrides marked `kind: restate` for clarity: `headline.seo_title_char_count` (faq/recipe), `headline.numerals_in_heds` (WW), `grammar.couple_verb_agreement` (WW), `linking.internal_link_count` (faq). (5) **Three new register rules** added covering format/publication-specific behaviors: `format.timeline.chronological_order` (timeline H2 entries must be in chronological order—earliest first), `publication.us-weekly.brand_italicization` (italicize *Us* and *Us Weekly* in body), `publication.us-weekly.tv_film_title_italics` (single-quoted Title-Cased phrases in USW body flagged for italics conversion). **Prose-level**: the recurring misattribution that "USW deviates from AP on em-dashes" or "WW matches AP" was corrected across all six docs that carried it—real AP uses spaces, USW matches AP, WW is the AMI deviator. **Total**: 1 P0 data-bug fix + 5 doc-misattribution corrections + 5 register cleanup items + 3 new rules. Driver: comprehensive audit pass on `write-assist/docs/js/engine/` revealed the inversion; canonical AP-Compatible-quick.md + ap-compatible-condensed.md sections cited as authoritative. Human-readable `docs/conflict-register.md` twin updated to mirror all YAML changes. ## [1.9.9]—2026-05 - **Us Weekly + Woman's World gap-fill from admin-panel voice-guide audit.** Cross-checked the CSA admin-panel "Style Guides for Pierce Audit" document (~/Downloads/style-guides-for-pierce-audit.2.md) against canonical `docs/us-weekly.md`, `docs/womans-world.md`, and the three `docs/ap-compatible-*.md` tiers. **AP-Quick** is byte-identical to admin and **AP-Condensed** / **AP-Thorough** canonical are strict supersets of the admin equivalents → no changes needed. **Woman's World** canonical is already a strict superset; one micro-clarification added (recipe substitution suggestions framed for common dietary needs). **Us Weekly** canonical was missing nine production rules the admin guide carries: (1) `*Us*` italicized with capital U whenever the word "us" appears as a brand nod; (2) TV / film titles use single quotes in heds + italics in body (split rule made explicit); (3) months always spelled out—never abbreviated, AP override; (4) date format depends on recency—four-row context table (today-this-week / earlier-this-year / previous-year / evergreen-quote); (5) first-and-last-name reference on first occurrence within each H2 section; (6) quote integrity—reproduce exactly, no mid-sentence `…` truncation, preserve censored-asterisk profanity verbatim; (7) H2 prose-paragraph article structure required, "Key Facts" bullet format banned; (8) headlines 80–100 characters with single-theme requirement on roundups; (9) word-count adherence with ~15% overage cap. Two new top-level sections added: `### Evergreen vs. News Article Handling` and `### Word Count`. Top-of-page red-overrides summary, Pre-Publish Checklist, and What-to-Avoid table extended to surface all nine new rules. File grew from 353 → 442 lines. ## [1.9.8]—2026-05 - **`data-universe-labeling.md` cross-syndication distortion screen + TRACKER_ENRICHED caveat.** Two additions complying with the 4/20 universe-labeling standing rule. (1) New caveat on the TRACKER_ENRICHED universe entry: performance-signal columns (`is_hit`, `article_vs_co_median`, `cluster_vs_co_median`, `author_hit_*`) are origin-PVs basis post-screen (gated on the Marfeel→Snowflake feed lands); reach + revenue columns unchanged. Driven by exec/leadership 2026-05-06 ask (Field-Level-Media 24-site example). (2) New universe entry for `MARFEEL_ARTICLE_BY_MEDIUM` describing the cross-syndication distortion screen—separates origin from syndicated PVs, tags articles with `SYNDICATION_JUICE ∈ {none, light, heavy}`, fail-open NULL handling. Distinguished from existing `STORY_TRAFFIC_METRICS` (CROSS_SITE_*) which covers internal McClatchy syndication only; one-line clarification added to that entry to prevent conflation. Implementation cross-linked to ops-hub master spec, data-headlines UI spec, data-keywords pre-flight note. ## [1.9.7]—2026-05 - **Site-wide consistency + completeness pass.** Audit ran across every `docs/*.md` page, every `_data/rules/*.yaml` artifact, every cross-reference, and every entry in `api/reference.json`. Issues fixed: `docs/changelog.md` + `docs/documentation.md` + `docs/data-universe-labeling.md` were each missing one of the AGENT-AUDIENCE machine-grep hooks (frontmatter `agent_audiences` field on the first two; in-body `` HTML comment on the third). All three now have both. **Navigation:** Layered Enforcement nav-pane position moved out of the top-of-nav editorial section and into the operational section under Personas (already shipped in the prior commit; this entry catches it in the changelog). Two new dividers added in the lower nav grouping: Publishing Guidelines (publishing workflow) | Acceptable Sources / AI Tool Responsibility / Claims Validation / Data Universe Labeling (editorial integrity standards) | Follow-Up Content (editorial strategy). **Verified:** all 39 `docs/*.md` pages have working `[⬇ Download]` buttons resolving to same-origin `/assets/sources/.md`; assets/sources/ contents have parity with docs/ (39 files each); `_data/rules/*.yaml` (schema + precedence + conflict-register) all parse cleanly; master-reference §1–§11 in numeric order with `` markers per section (11 markers, 11 top-level sections); version stamps consistent (master-reference top-of-page = footer = api/reference.json `version` field = 1.9.7); all internal `/docs/` links resolve to actual docs/ files; api/reference.json `tag_definitions_url` + `raw_markdown_url` + per-format `standalone_url` all resolve; `_data/rules/*.yaml` published to `/assets/sources/rules/` for engineering-side ingest. **Changelog now reflects every edit shipped in this session** (v1.9.1 → v1.9.7). ## [1.9.6]—2026-05 - New §11: **Layered Enforcement**—conceptual model + machine-readable rules contract for layered style guidance enforcement, addressing engineering's documented confusion about how General / Persona / Article Format / Platform layers should compose + how conflicts resolve. Three deliverables shipped: **(1) Concept doc** at `docs/layered-enforcement.md` explaining the four-layer model, override-vs-stacking semantics, six worked examples (em dash on a USW ETK, couples-verb on a WW cast piece, headline char count on an FAQ-on-Apple-News, etc.), reconciliation with CSA backend's documented Trust Hierarchy (per engineering's PGS Confluence "How Style Guides Work" doc, page 1949663238), proposed Composer extension, implementation routing (writer / grader / composer / human enforcement targets), and what-this-doesn't-solve. **(2) Machine-readable rules contract** at `_data/rules/`: `schema.yaml` (rule-entry shape including domains, value types, selector grammar, enforcement targets), `precedence.yaml` (resolution algorithm with tie-breakers—most-specific-selector → most-recent-introduction → data-validated-wins → editor-resolution; null-override + additive-only-rule + per-rule precedence-override special cases; Composer extension proposal; enforcement routing; validation timing). `conflict-register.yaml` with **48 rules** spanning headline (16) / body (5) / punctuation (3) / grammar (2) / structure (3) / voice-tone (8) / compliance (8) / linking (2) / media (4) / cms (1) / faith (1) / format-specific (4) / cross-cutting explicit-language (2). All cross-publication divergences captured (USW plural-couples + space-em-dashes; WW Dolly Parton tone north star + meno-brain rationale; TH B2C psychographic-only). **(3) Human-readable register** at `docs/conflict-register.md`—hand-maintained Markdown twin with grouped tables + a worked-example roll-up showing how `{ persona: curious-optimizer, format: faq, platform: apple-news }` resolves across 14 rules. **Critical engineering reconciliation surfaced:** the CSA backend currently STACKS layers (no override semantics) per its own Confluence doc—fundamental semantic gap with the editorial 4-layer model. Persona + Format are not separate layers in the current Composer; `style_guide` field is informational-only; National Team users get no platform-specific voice. The §11.5 reconciliation section makes the gap explicit + proposes a Composer extension. **Infrastructure:** `scripts/build-downloads.sh` extended to publish `_data/rules/*.yaml` to `/assets/sources/rules/` so engineering can ingest the YAML directly. Master-reference §11 condensed summary added; §10.6 AP-Compatible block reordered to come before §11; navigation, documentation, index, api/reference.json updated (new section + per-section `rules_artifacts` + `companion_pages` keys). ## [1.9.5]—2026-05 - New §10.6: **AP-Compatible** reference standard added as a four-page tiered structure—an overview at `docs/ap-compatible.md` plus three nested tier pages at `docs/ap-compatible-quick.md` (~2K tokens / ~1,600 words), `docs/ap-compatible-condensed.md` (~12K tokens / ~9,600 words; recommended default), and `docs/ap-compatible-thorough.md` (~25K tokens / ~19,000 words). All four slugs match McClatchy CSA backend `style_guides/sources/` convention; the three tier pages are direct wrappings of the existing CSA backend seed files (Jekyll frontmatter + Back-to-overview link + download button + tier-context intro added; body content unchanged for zero-friction handoff). The overview page (~3K tokens) provides: the three-tier explanation table; direct download links per tier; **when-to-use guidance** cross-linking to outlet-specific style guides (US Weekly §10.3, Trend Hunter B2C §10.4, Woman's World §10.5) and distribution platforms (Apple News §10.2, SmartNews §10.1); **cross-publication divergence table** flagging the three rules that diverge meaningfully across McClatchy outlets (couples / pairs / duos verb agreement; em dash spacing; Oxford comma); headline rules quick reference; reference hierarchy. The Quick tier covers 21 flat sections (essentials); Condensed covers Parts 1–12 (Punctuation / Capitalization / Numbers / Abbreviations / Titles / Attribution / Word Usage / Race & Identity / Legal / Sports / Business / Datelines); Thorough covers Parts 1–22 (adds Weather / Health / Science & Tech / Religion / Elections / Military / Food / Environment / Transportation / Headlines & Captions) plus appendices (commonly misspelled words / redundancies / clichés / problematic phrases / tricky plurals). The three tiers are nested—Quick ⊂ Condensed ⊂ Thorough—so picking the right tier means picking the right depth, not picking different content. Master-reference §10.6 condensed summary added; navigation, documentation, index, api/reference.json updated (api `ap-compatible` entry includes nested `tiers` array). ## [1.9.4]—2026-05 - New §10.5: **Woman's World** publication style guide added as `docs/womans-world.md` (slug matches McClatchy CSA backend `style_guides/sources/` seed-file convention). Source consolidation: existing CSA backend seed file (`backend/style_guides/sources/womans-world.md`, ~360 lines / ~2,200 words) merged with two newly-delivered PDFs from content team lead—`brandguide_WW.pdf` (the AMI Women's Group Digital Style Guide, ~90% identical to Us Weekly's Master Style Guide) and `styleguide_WW.pdf` (the brand voice doc—mission, target demo, tone celebrities, USE/AVOID lists with rationales, 15-item Health Tentpoles, brand notes). The PDF filenames are label-swapped vs. their actual contents. **Critical cross-publication divergence surfaced:** WW uses **singular verbs** for couples / pairs / duos (mashups always plural), the AMI default—**opposite of Us Weekly**. WW also uses **no-space em dashes**—an AMI Women's Group house-style deviation from AP (which uses spaces; ap-compatible-quick.md + ap-compatible-condensed.md), opposite of Us Weekly's AP-conformant space-em-dash convention. Page covers: audience persona (Gen X 50s through Boomers 60s+; Dolly Parton tone north star); voice and tone (Warm / Empathetic / Optimistic / Practical / Trustworthy; "from your best friend" approach); USE/AVOID word lists with rationales (incl. "meno brain" stigma reasoning); four writing personas (rotate, never use the same on every article); seven topic pillars (Health & Wellness with 15 tentpoles · Beauty & Fashion · Food, Cooking & Recipes · Books & Entertainment · Faith & Values · Holidays & Seasonal · Home / Lifestyle / Self-Improvement / Consumer Advice / Feel-Good Stories); brand notes (psychographic over demographic; treat-vs-tool; promising happy endings; money-conscious; many roles; hard topics with positive frame; "you're not alone"); style reference (key WW-vs-USW differences + shared AMI conventions); WW-specific quality rules (5-year-recent peer-reviewed health claims; recipes that work; products under $20); what-to-avoid table; pre-publish checklist. ~3,350 words / ~4,500 tokens—within Layer 2 voice-guide cap (≤5K tokens). Master-reference §10.5 condensed summary added; navigation, documentation, index, api/reference.json updated. ## [1.9.3]—2026-05 - New §10.4: **Trend Hunter B2C** publication style guide added as `docs/trend-hunter-b2c.md` (slug matches McClatchy CSA backend `style_guides/sources/` seed-file convention). Source: existing CSA backend seed file at `backend/style_guides/sources/trend-hunter-b2c.md` audited and confirmed already in proper writer's-bible format—minimal transformation needed beyond formatting for the docs site (frontmatter, agent-audience comments, red-span overrides, download link, cross-link to §4.2 Curious Optimizer persona). Audience: **The Curious Optimizer**—psychographic persona spanning all demographics. Voice: Smart, Approachable, Curious, Practical. Four reader intent tiers (Discovery / Understanding / Evaluation / Action). Article structure: What It Is / Why It Matters / Who It's For / How to Experience It. Short paragraphs (2–3 sentences max), bulleted lists, pull-out Quick Takeaways. Strongest anti-pattern: **demographic stereotyping in any form**—frame benefits psychographically ("If you want…", "For anyone curious about…"), never demographically. Banned trendy slang (slay / bestie / vibes), academic density, clickbait. ~1,300 words / ~1,700 tokens—well within Layer 2 voice-guide cap. Master-reference §10.4 condensed summary added; navigation, documentation, index, and api/reference.json updated. ## [1.9.2]—2026-05 - Section §10 renamed: **Platform Formats → Platform Guidance**. Scope broadened from distribution-platform requirements only to encompass per-publication style guides as well. Updated in `_data/navigation.yml`, `docs/master-reference.md` (heading + SECTION marker + scope text + cross-link), `docs/documentation.md`, `docs/index.md`, `api/reference.json` (id `platform-formats` → `platform-guidance`, URL anchor, notes), and cross-references in `docs/headlines.md` (×2) + `docs/publishing-guidelines.md` (×2). Existing platform pages (Apple News, SmartNews) and their slugs unchanged. - New §10.3: **Us Weekly** publication style guide added as `docs/us-weekly.md` (slug matches McClatchy CSA backend `style_guides/sources/` seed-file convention for zero-friction handoff). Writer's-bible format covering: audience persona (inferred + flagged for editor sign-off), voice and tone, headlines (sister-rule across onpage / promo / SEO with shared keywords; numerals not spelled-out; single quotes; no periods / ampersands / Oxford commas), H2 subheads (title case + keywords + complete phrases / questions), body conventions (bold celebrity names + tag-page links; em dashes and ellipses with spaces on both sides; numbers spelled out 1–9; italics rules; closed-up `co-` compounds; commas; couples / pairs / duos take plural verbs—Us Weekly only; curse words; apostrophe-s rule), curated A–Z style reference (high-frequency Us Weekly–distinctive entries; full Master Style Guide remains source of truth), grammar and punctuation, CMS workflow, reference hierarchy (Us Weekly Style Supplement → Merriam-Webster's Collegiate Dictionary 11th Edition → AP Stylebook → Chicago Manual of Style → Words Into Type), what to avoid table, pre-publish checklist. Sized for Layer 2 voice-guide cap (≤5K tokens) per CSA backend trust hierarchy. Master-reference §10.3 condensed summary added. ## [1.9.1]—2026-05 - Article Formats §3.11 (FAQ / Service Journalism): per-block word count tightened from 300–500 to 100–200 words per question/answer block. Tighter blocks read more snippet-ready and reduce padding pressure. Updated in `docs/faq.md` (override summary, Word Count section, Pre-Publish Checklist) and `master-reference.md` §3.11. No structural changes; H2 count, answer-first rule, and 2–3 sentence paragraph max unchanged. Requested by content team lead 2026-05-08. ## [1.9.0]—2026-05 - Article Formats §3: new format added—§3.12 What to Know Next. Forward-looking explainer for The Curious Optimizer (§4.2): three-section arc (what's happening, why, what could be next) closing on a reader-actionable takeaway. 600–900 words; mandatory three-section proportionality; sourced causation in §2; paired prediction-plus-action in §3; forward-looking signal required in headline. Disambiguates from Discover Explainer (informational/static) and Recap (retrospective). - Infrastructure sync: navigation.yml (new live entry inserted alphabetically after Google Discover Explainer in the non-pending group); `_includes/headline-formulas/what-to-know-next.html` (accordion partial for the Headlines §2 Format-Specific block—missing partial broke the Cloudflare Pages build for commit 5497abd); api/reference.json (version 1.9.0, new format entry); documentation.md (format count 11 → 12, new row in Active Article Formats); index.md (format count 11 → 12); master-reference.md (§3.12 condensed summary). ## [1.8.3]—2026-04 - Subheading (H2) Keyword Rules added to General Guidelines §1.2 and mirrored in `master-reference.md` §1.2. Rule: primary keyword must appear in at least one H2; secondary keywords should appear in other H2s when they fit naturally (don't force). Surfaced by content team feedback on just-shipped keyword enforcement (PGS-104)—thorough in on-page title, SEO title, meta description, and intro but not enforced in H2s—and confirmed by content team lead. Tool-side enforcement gap to be ticketed as follow-up by product leads; content-standards side documents the rule regardless of current tool coverage. ## [1.8.2]—2026-04 - "Variants must be ≥50% different" blanket claim retracted site-wide (2026-04-24). Semantic similarity floors at ~88% by construction for fact-identical variants (same draft → same facts), so a blanket 50% different claim measures the wrong layer or was wrong. Replaced across `brand-guidelines.md`, `tool-responsibility.md` (×4 references, including persistent-issue escalation language), `master-reference.md` (×3), and `discover-explainer.md` (×2) with: "meaningfully differentiated in structure and prose." New language recognizes the 4-dimension measurement model (semantic / structural / lexical / surface) from the differentiation prototype handoff 2026-04-24. Specific thresholds are under calibration—tracked as P23 Content Differentiation Framework. **Kept:** all other variant requirements (local + national variants, revival optimization, peer review, fact verification). Handoff doc is v0.2 prototype; thresholds land after content team lead format-only 2-day test + internal consensus. ## [1.8.1]—2026-04 - General intro word-count rule retracted site-wide (2026-04-23). SEO team broadly retracted the 80–120 word intro length rule. Tool-side enforcement dropped (PGS-147, PGS-205 → WON'T DO). Standards-side updates: removed "minimum H2 intro length (80 words)" framing from everything-to-know.md (override summary, body sections ×3, Formatting Rules item, Pre-Publish Checklist line, What to Avoid violations table); removed "Gallery or listicle intro 80–150 words" + "Breaking news jam 80 words minimum" entries from headlines.md word-count table; removed "80-word intro minimum" annotations from master-reference.md H2 template entries. content strategist correction noted: the everything-to-know.md framing applied "every H2 section," but the rule (when active) was article-intro-only—fix surfaced during retraction. **Kept:** headlines.md "first 80 words for keyword placement" framing—different concept (keyword placement, not minimum word count) and content strategist explicitly preserved as soft guideline. ## [1.8.0]—2026-04 - Article Formats §3 / Personas §4: Restructured persona-adjacent format pages per updated CSA product architecture. content team lead confirmed that format guidance per persona can be dropped once personas and formats are decoupled in the CSA UI—format selection handles format; persona handles audience. Removed §3.12 Trend Feature, §3.13 Science & Discovery Feature, and §3.14 Entertainment & Trending as standalone format pages. - Persona pages updated: §4.2–§4.5 each gained a **### Tone** section preserving the tone guidance that was persona-level (not format-level). Cross-references to deleted format pages removed from CSA Target Audience Definition sections. All other persona content (Who They Are, What They Respond To, Content Framework, Content Implications, CSA Target Audience Definition) unchanged. - Infrastructure sync: navigation.yml (three active format entries removed), admin/config.yml (three CMS entries removed), api/reference.json (version 1.8.0, three format entries removed), documentation.md (format count 14 → 11, three rows removed from Active Article Formats table), index.md (format count 14 → 11), master-reference.md (§3.12–3.14 sections removed, §4.2–4.5 Format page cross-reference lines removed), three headline-formula includes deleted (trend-feature.html, science-discovery-feature.html, entertainment-viral.html) ## [1.7.0]—2026-04 - Personas §4: CSA Target Audience Definition section added to all five persona pages—formatted definition for direct entry into the CSA product UI (Name, Description, Focus areas: Discovery / Understanding / Evaluation / Action). Section demarcated with horizontal rule and blockquote callout. AGENT-AUDIENCE: general-style tagged. Each section cross-references the relevant format page. - Article Formats §3: Three new format pages added—§3.12 Trend Feature, §3.13 Science & Discovery Feature, §3.14 Entertainment & Trending. Each follows the full format page apparatus: AGENT-AUDIENCE section comments, frontmatter agent_audiences, download buttons, Content Framework, Tone, Writing Guidance, What to Avoid, Pre-Publish Checklist. No locked headline formula or CSA product-level specs yet specified—general-style and editorial writing guidance only. - §3.12 Trend Feature: primary persona The Curious Optimizer (§4.2); Content Framework Discovery / Understanding / Evaluation / Action; psychographic-first universal language required; no demographic assumptions - §3.13 Science & Discovery Feature: primary personas The Wonder-Driven Science Enthusiast (§4.3) and The Curious Explorer (§4.4); two audience modes documented (broad awe-driven vs. shareable-fact-driven); Content Framework Discovery / Understanding / Evaluation / Action - §3.14 Entertainment & Trending: primary persona The Watercooler Insider (§4.5); Content Framework Discovery / Understanding / Evaluation / Action; debunk and "real story behind" as highest-performing pattern - Persona pages updated: §4.2–§4.5 now include **Format page** cross-reference line pointing to the relevant new §3 page; master-reference §4 entries updated to match - Infrastructure sync: navigation.yml (three new format children under Article Formats), admin/config.yml (three new CMS entries with format-specific hints), api/reference.json (version 1.7.0, three new format entries), documentation.md (format count 11 → 14, three new rows in Active Article Formats table, §4 row updated to describe CSA Target Audience Definition sections), index.md (format count 11 → 14), master-reference.md (§3.12–3.14 condensed summaries added, §4 scope updated, all persona entries updated with format page cross-reference), three new headline-formulas includes (trend-feature.html, science-discovery-feature.html, entertainment-viral.html) ## [1.6.3]—2026-04 - Headlines §2 (Us Weekly): removed "H2 post / explainer intro—80 words minimum" word count requirement; dropped by L&E/UsW team after review with TEO (data team) ## [1.6.2]—2026-04 - SEO Keywords (Meta Keywords) guidance added to General Guidelines §1.2: new `### SEO Keywords (Meta Keywords) Rules` subsection; field added to §1.2 Character Counts; SEO keywords line added to §3.1 Pre-Publish Checklist - SEO Keywords field added to all ten §3 article format sections (§3.1–§3.10): `### SEO Keywords (Meta Keywords)` **(REQUIRED)** subsection inserted after `### Meta Description` in each format—1–5 keywords, single words acceptable, all lowercase, comma-separated, location names required for local stories ## [1.6.1]—2026-04 - Platform Formats consistency audit and override validation: all rules from validated platform format guidance confirmed captured; all discrepancies corrected; data-validated claims labeled throughout - Publishing Guidelines §6.3 Apple News: headline range corrected from stale 80–139 (flatplan.io source) to validated 90–120 (110–119 sweet spot); no-question and no-number-led rules added; separate Apple News title requirement added; subtitle required added; defers to §10.2 as authoritative spec; stale flatplan.io attribution removed - Headlines §2: Apple News block under Us Weekly rewritten—separate title requirement (90–120 chars) replaces stale "inherits UsW counts" and "no Apple News-specific formula" language; no-question/no-number-led rules added and marked red; featured placement exception documented; cross-reference to §10.2 - Headlines §2: SmartNews outlet block added—70–90 chars (80–99 sweet spot, data-validated), formula constraints (no questions, no WTK, number-led positive), RSS `