agentsclimarketplace

New blog

Skill popiliadam/platinum-seo-engine/skills/production/new-blog

Use when: kullanıcı "yeni blog yaz", "blog üret", "content gen", "blog draft", "makale yaz", "yeni içerik üret", "new blog post", "pillar yaz", "cluster yaz" der ya da /pseo-new-blog çağırır; master.xlsx[new_content_plan] satırından plan al, SERP analiz yap, 5 template render et, JSON-LD @graph 5 entity üret, Foundational Principles 3-katman (truth-verifiable + profile-aware + anti-cheap-content) enforce et. Also use when: master.xlsx[topical_map] içinde assigned_url=null olan pillar/cluster/supporting satırı için ilk içerik üretilecek; new_content_plan satırının lifecycle_status=GREEN ya da PLANNED durumda; mode='draft' (review) veya 'publish' (immediate) seçildi; profile-aware switch tetik (project-config[profile] ya da project-config[profiles] priority merge). Do not use when: mevcut blog revize için (revise-content kullan); FAQ block re-render için (faq-optimization kullan); decay/sunset/prune için (content-remediation kullan); image generation için (generate-images kullan); manuel WordPress upload için (Section A upload-instructions template render edilir, skill upload yapmaz); master.xlsx eksikse (init-project önce çalışmalı, DURUR #1).From its SKILL.md

Install
npx -y skills add popiliadam/platinum-seo-engine --skill new-blog

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

34.6 KB, as published. Nobody here has run it

new-blog — production skill (Phase 11 Wave 1, W-F1)

Plan-driven full-article generator. Reads a single master.xlsx[new_content_plan] row, joins it against master.xlsx[cluster_keywords] (keyword set + intent + forbidden filters) and master.xlsx[topical_map] (pillar/cluster/supporting page_type + internal-link plan), runs SERP top-10 + mandatory Tier-1 Scrapling competitor recon (×10 organic, CCE arming) and gates the draft against the Brief Paketi, then renders four artifacts:

  1. outputs/blog/{slug}/article.html — <article class="pse-blog-post"> fragment (R-22 fragment boundary; no <html> / <body> wrap).
  2. outputs/blog/{slug}/schema.jsonld — JSON-LD @graph with 5 entities (Article + Organization + Person + BreadcrumbList + FAQPage; R-78..R-83).
  3. outputs/blog/{slug}/meta-tags.json — meta title (≤580px desktop)
    • meta description (≤990px desktop) per R-35 (single authority) + OpenGraph + Twitter card pack.
  4. outputs/blog/{slug}/upload-instructions.md — Section A new-blog manuel WordPress upload workflow (R-74); skill itself does NOT publish.

The skill is READ-ONLY with respect to master.xlsx (new_content_plan allowed_writers is null — F-1 schema authority, verified pre-flight). It never calls transaction.append, transaction.update, or transaction.delete against the workbook. The only state mutation is the audit append to _state/events.jsonl with event_type=content_new.

Inputs (frontmatter contract)

NameTypeDefaultNotes
project_slugstring—Required. projects/{slug}/ workspace root.
new_content_plan_idstring—Required. Row id (kolon A) in master.xlsx[new_content_plan].
modestringdraftdraft (review) or publish (immediate). Other values → DURUR #5.

Outputs (artifacts produced)

  • outputs/blog/{slug}/article.html — schema-markup-aware, WCAG 2.1 AA compliant, BEM pse- CSS class prefix HTML article fragment.
  • outputs/blog/{slug}/schema.jsonld — JSON-LD @graph 5 entity (Article + Organization + Person + BreadcrumbList + FAQPage).
  • outputs/blog/{slug}/meta-tags.json — meta title (≤580px) + meta description (≤990px) per R-35 + OG + Twitter pack.
  • outputs/blog/{slug}/upload-instructions.md — Section A new-blog WordPress manual upload steps (R-74).
  • _state/events.jsonl — audit append, event_type=content_new, event_kind=work (events.schema enum: provenance / work / audit / workflow), schema_version=1.0.

Foundational Principles Enforcement (3-Layer)

The 3 üst-prensip (Phase 10 rules/content-quality.md#foundational-principles) gate this skill end-to-end. No alt-rule (R-01..R-148) overrides them.

Principle 1 — Truth-Verifiable Content (R-27, Süleyman 5x vurgu)

Statement. Tüm content/source/link/data %100 doğru ve kanıtlanabilir; uydurma yasak (kaynak, hikaye, case study, fiyat, stat, ürün/feature/ image-link).

3-katman defense (Step 8 implements):

  • Layer 1: skill prompt explicit "uydurma yasak" başlangıçta — Step 7 generation öncesi system note inject.
  • Layer 2: post-generate fact-check pass — her claim için content_settings.original_research_database + SERP top-10 source match; citation count ≥ R-44 threshold.
  • Layer 3: citation enforce — R-44 source verification 3-katman + R-105 expert quote bank-driven (content_settings.experience_database lookup, uydurma quote yasak).

Failure mode: RED — yayın iptal, çıktı discard. DURUR #6 trigger.

Principle 2 — Profile-Aware Enforcement

Statement. Skill behavior project.config.json[profile] enum'una göre değişir (singular field, Phase 11 W-F1 cascade fix). When profile is unset, fall back to profiles[] priority merge per rules/content-quality.md Principle 2 tablosu.

Enum 5-value: e-commerce / ymyl / local-service / b2b-saas / portfolio.

Switch logic (Step 2 + Step 7 + Step 8):

BoyutYMYLe-commerceb2b-saaslocal-serviceportfolio
Author byline (R-28)ZorunluYokEsnekEsnekYok
Tonesemi-pro + sizconversational+senformal + sizconversational+sizesnek
Word count1500–4000800–25001800–35001000–2500800–2000
Counter-argument (R-50)ZorunluSkipOpsiyonelSkipSkip
Disclaimer (R-51)medical/legal/financialSkipOpsiyonelSkipSkip
Image styleclean-illustrationproduct-photodiagram-screenshotlocation-photoesnek
Stats density (R-104)min 1/500w, max 1/200wmin 1/800w, max 1/300wmin 1/600w, max 1/250wmin 1/1000w, max esnekesnek

Failure mode: AMBER — wrong profile resolution worker raporlar; RED ancak profile enum dışı (DURUR #2).

Principle 3 — AI Suistimal Önlemi (Anti-Cheap-Content, 7-pattern gate)

Statement. AI'ın doğal cheap content padding davranışını preempt et.

7-pattern gate (Step 8 implements):

  • H3 zorunluluk gate — H2 word count > 200 → min 2 H3 (R-30).
  • Heading keyword density %40–60 — primary_keyword count / total heading count (hard cap, R-30 stuffing önleme).
  • Citation density per 500 word 1-2 (R-106).
  • FAQ count talep-güdümlü 3-6 (kanıt varsa), hard cap 10 (R-09).
  • Stats density profile-aware min/max (R-104 tek kaynak / single source; yukarıdaki tablo R-104'ün min+max cap'lerini birebir taşır, new-blog kendi sayısını uydurmaz — FIX-H5 hizalaması).
  • Per-H2 list cap 1 (R-07; multi-list AI padding önleme).
  • AI signature humanize — brand_identity.tone_phrases_blocklist consume + replace pattern ("Aslında", "Sonuç olarak", "Özetle", R-118).

Failure mode: AMBER warning (auto-correct attempt) → RED fail (manuel revise). 2x AMBER aynı pass → RED upgrade.

R-121 Bank Selection Logic (Pre-Publish)

Bank entry (R-105 expert quote / R-114 original research / R-119 first-hand experience) selection passes a 3-step filter before any template renders. Step 7 (render_5_templates) reads the result; Step 12 (emit_provenance) records the chosen entries. R-121 catches semantic repetition (the same core fact copy-pasted across many blogs); R-118 catches stylistic repetition (AI signature words within one page). The two rules compose — both fire pre-publish.

  1. Topic match. entry.applicable_topics ∩ (new_content_plan.primary_keyword ∪ new_content_plan.assigned_cluster ∪ topical_map[matching_row].cluster) ≠ ∅ → entry remains a candidate; empty intersection → entry skipped.
  2. Profile density cap. From the candidate set, select up to the per-profile maximum:
    • YMYL: 2 experience + 1 research entry per content
    • b2b-saas: 1 experience + 1 research entry per content
    • e-commerce / local-service / portfolio: 1 experience + 0 research entry per content
  3. Rotation (30-day). For each remaining candidate, count usage in the append-only _state/events.jsonl content_new work events whose note records bank_entry_id == entry.id AND timestamp >= now - 30d. Skip if count >= entry.max_usage_per_month; rotate to a different entry or to a different phrasings[] form of the same entry.

If all candidates fail the 3-step filter, the skill emits AMBER and retries with phrasings[] rotation; 2x AMBER same pass → RED upgrade (manual review).

Post-publish usage record (in the events.jsonl audit append at Step 12): the content_new work event records the bank_entry_id(s) used inside its note field — the append-only _state/events.jsonl is the only state surface this skill writes (consistent with the READ-ONLY claim above). The rotation counter at Step 3 reads from this same append-only content_new event log on the next invocation — single source of truth. The skill does NOT write master.xlsx[completed_work] (allowed_writers=null — F-1); if a durable completed_work row is later required, the done-protocol / mark-done skill owns that append.

Schema enablement (v1.4, commit 8e07e1c): each bank entry exposes applicable_topics, phrasings, last_used_in_content_id, max_usage_per_month. Stage C of brand-onboarding (commit cb8df43) populates these via R-44 evidence-gated atomic write.

Runtime integration deferred to Phase 11 Wave 2. This SKILL.md section is the spec lock. scripts/production/new_blog.py now exists (CCE B4 orchestrator helpers assemble_brief / evaluate) but does NOT yet implement the R-121 3-step bank filter. When Wave 2 lands that runtime, the 3-step filter implements as a pre-render_5_templates helper consumed by Step 7.

R-124 — YMYL Uzman İnceleme İmzası (Pre-Publish Expert-Review Sign-Off)

YMYL profillerinde (profile == "ymyl"), R-124 gereği yazar künyesi (R-28) yalnızca kayıtlı bir insan incelemesi ile geçerlidir. Bu skill, YMYL içeriğini yayınlamadan ÖNCE zorunlu bir inceleme-imzası adımı çalıştırır (Step 11.5 ymyl_review_signoff, Step 11 ile Step 12 arasında):

  1. Operatör inceleyiciyi adlandırır. Yayın öncesi, operatör (Süleyman ya da atanmış uzman) insan reviewer (inceleyenin) adını verir. Reviewer adı yoksa → DURUR #9 (RED, yayın bloklu — R-124).
  2. Audit kaydı (events.jsonl, append-only). Skill, inceleme kanıtını scripts/state/events_writer.append_audit(...) ile bir event_kind=audit satırına yazar:
    • audit_action="modified" (events.schema 6-değer canonical enum {created, modified, deleted, accessed, permission_changed, config_changed} — içeriğin yayın-öncesi inceleme kapısı onaylandı/değiştirildi).
    • audit_target="content:{project_slug}:{post-slug}".
    • actor="skill:new-blog".
    • notes payload üç zorunlu alan taşır (R-124): reviewer (inceleyen adı) + review_date (ISO 8601) + content_version (content hash veya R-103 revision id).
  3. Eşleştirme. Bu audit satırı, R-28 künyesinin arkasındaki gerçek insan inceleme kanıtıdır (Principle 1 Truth-Verifiable + EEAT Trustworthiness). İnceleme kaydı olmadan adlı yazar künyesi = uydurulmuş yazarlık sinyali (Google rater-guideline "Lowest"). R-28 künyeyi görünür kılar; R-124 künyenin arkasında belgelenmiş bir insan inceleme olmasını zorunlu kılar.

Non-YMYL profillerde bu adım atlanır (R-28 künye opsiyonel/yok), audit satırı yazılmaz.

CCE Döngüsü (Competitive Content Engine — Silahlan → Yaz → Kapı → Re-write)

new-blog artık CCE motorunu çalıştırır: rakipten ölçülebilir biçimde daha iyi içerik üretmek için bir silahlan → yaz → kapı → re-write döngüsü. Mimari kural (spec §3): döngüyü Claude (bu skill) sürer çünkü scripts/production/new_blog.py SAF Python'dur — MCP'yi/Claude'u ÇAĞIRMAZ; döngüye yalnızca iki saf yardımcı verir (assemble_brief, evaluate). MCP sınırı skill'e (Claude'a) aittir; deterministik tutkal koda.

1. Silahlan (Arming) — ham MCP JSON → Brief Paketi

Claude ham veriyi MCP ile toplar (kod değil), diske yazar, sonra new_blog.assemble_brief(...)'i çağırır:

SinyalMCP tool (Claude çağırır)Not
SERP top-10 organicmcp__dataforseo__serp_organic_live_advanced depth=10rakip sıralaması + URL listesi
Rakip DOM ×10mcp__ScraplingServer__stealthy_fetchZORUNLU her organic URL (Step 4)
Keyword genişlememcp__dataforseo__dataforseo_labs_google_keyword_suggestions + …_keyword_overviewreliable substring + volume/intent
AI Overviewserp_organic_live_advanced ai_overview bloğugerçek answer_points (Step 5)
GSC sinyalimcp__gsc__search_analytics{real_queries, current_position}

assemble_brief 3 B1 transform'u sırayla çalıştırır (competitor_recon_transform

  • keyword_aio_intel_transform + build_brief) → tam Brief Paketi (gap + structure_ceiling + keyword_set + aio + gsc). GSC sinyali gsc= parametresiyle build_brief'in boş default'unu override eder.

2. Yaz (Craft) — Claude, R-149+ rehberiyle

Claude makaleyi yazar; rules/content-craft-discipline.md (R-149..R-153) pozitif yazım katmanını uygular: pain-mirror empati girişi (R-149), somut örnek/senaryo (R-150), aktif marka sesi (R-151), rakip-üstü özgün değer data-original="true" (R-152), H2 derinliği + mantıksal akış (R-153). Brief'in gap maddeleri kapsanır; structure_ceiling+1 anlamlı tablo/liste hedeflenir. Foundational Principles 3-katman + R-121 bank seçimi yine geçerlidir (üstte).

3. Kapı (Gate) — new_blog.evaluate(...)

Üretilen HTML new_blog.evaluate(content_html, brief, round_num=..., max_rounds=3) ile ölçülür; bu scripts/production/quality_gates.py'nin run_all'ini IMPORT ile çağırır (subprocess değil) — 8 kapı sırasıyla: p0_truthfulness > gap > structure > depth > aeo > voice > originality > schema_visual. Döner: {gate_result, action, feedback}.

4. Re-write / Terminal — döngü kararı

actionAnlamSonraki adım
PASSoverall PASSyayın akışına devam (Step 9+)
REWRITEoverall RED, round < 3feedback (= rewrite_feedback) ile YENİDEN yaz (adım 2)
AMBER_TERMINALoverall RED, round ≥ 33-tur cap — sonsuz döngü YOK; eksik kapılar operatöre raporlanır (spec §8)

AMBER yalnız orkestratör kararıdır; kapıların kendisi PASS/RED kalır. Döngü PASS'e ya da AMBER_TERMINAL'e ulaşana kadar adım 2 ↔ 3 tekrar eder. Craft kuralı ↔ kapı eşlemesi: R-149 → voice+aeo, R-150 → depth, R-151 → voice, R-152 → originality, R-153 → depth (content-craft-discipline.md).

Routing — 12-Step Workflow

Step names are stable identifiers across runs (used as steps[*].name when the skill calls workflow_runner.create_run).

Step 1 — read_master_xlsx

master.xlsx[new_content_plan] row read where id == {input.new_content_plan_id}. master.xlsx[cluster_keywords] filter where assigned_url == new_content_plan.url_slug. master.xlsx[topical_map] filter where pillar == new_content_plan.assigned_cluster OR cluster == new_content_plan.assigned_cluster.

Workbook opens with openpyxl.load_workbook(read_only=True). F-1 allowed_writers=null discipline: skill MUST NOT call transaction.append, transaction.update, or transaction.delete.

new_content_plan 14 columns consumed: id, title, url_slug, primary_keyword, monthly_volume, assigned_cluster, target_word_count, priority, created_date, tivl_tag, lifecycle_status, image_prompt, alt_text, content_type (Phase 10 additive bump). content_type enum: [listicle, guide, comparison, research, tutorial, review].

cluster_keywords 11 columns consumed: cluster, keyword, monthly_volume, data_source, assigned_url, gsc_clicks, gsc_impressions, gsc_position, intent, forbidden_kw, forbidden_reason. intent enum: [Informational, Commercial, Transactional, Navigational].

topical_map 10 columns consumed: pillar, cluster, primary_keyword, monthly_volume, data_source, assigned_url, page_type, status, priority, note. page_type enum: [pillar, cluster, supporting].

DURUR #1. master.xlsx eksik → init-project skill önce çalışmalı.

Step 2 — read_project_config (Profile-Aware Switch)

projects/{slug}/project.config.json parse. Singular profile field (schema v1.2, Phase 11 W-F1 cascade fix) is the dominant Principle 2 switch; when missing, fall back to profiles[] priority merge per rules/content-quality.md Principle 2 tablo.

content_settings 14 fields consumed (Phase 10): toc_strategy, related_posts_strategy, author_strategy, css_strategy, indexnow_enabled, ai_training_optin, video_integration, internal_data_sharing, external_uniqueness_check, original_research_database, experience_database, video_database, disclaimer_templates, image_model.

brand_identity 18 fields consumed (Phase 10): logo_url, primary_color, secondary_color, accent_color, font_family_heading, font_family_body, header_template_id, footer_template_id, source_url_for_sampling, tone, hitap, anglicism_tolerance, tone_phrases_blocklist, font_heading, font_body, default_hero_url, same_as_urls, image_style.

DURUR #2. Singular profile enum dışı bir değere sahipse RED; schema v1.2 cascade fix uygulanmadıysa (schema_version != "1.2" ve profile field schema'da yok) önce scripts/migrations/0002_project_ config_1.1_to_1.2.py çalışmalı.

Step 3 — serp_analysis (R-08)

mcp__dataforseo__serp_organic_live_advanced çağrı: keyword=new_content_plan.primary_keyword, depth=10. Top-10 sonuç + entity extract (organic results dom_url, title, description).

DURUR #5. SERP analiz fail (boş response, API error, mode enum dışı) → manuel input gerekli, skill awaiting_approval.

Step 4 — scrapling_competitor_recon (CCE silahlanma, R-08 derinleştirme)

ZORUNLU (artık optional değil; top-3 değil top-10). SERP top-10 organic sonucun HER BİRİ için mcp__ScraplingServer__stealthy_fetch ile DOM çekilir — CCE silahlanma motoru rakip kapsamasının TAMAMINI görmeli (spec §4a), yarısını değil. Her sayfa {"url": ..., "html": ...} olarak biriktirilir; bu liste + ham SERP assemble_brief'e verilir → competitor_recon_transform heading tree

  • answered questions + candidate entities + structure ceiling (anlamlı tablo/liste) + word_count çıkarır.

Tier-0 fallback (basic SERP description) yalnızca tek bir URL kalıcı 403/blok verirse o URL'e özgüdür; Scrapling required MCP bağımlılığı olduğundan adım topluca atlanamaz. Eksik/bloklu sayfalar AMBER warning raporlanır, silahlanma toplanan sayfalarla devam eder.

Step 5 — aio_citation_check (R-109/R-110/R-111 + CCE AEO)

Pasaj alıntılanabilirlik: SERP top-10 page'lerde featured snippet detect (R-109), people_also_ask (R-110), ai_overview citation (R-111).

CCE AEO bağı. Ham serp_organic_live_advanced payload'undaki ai_overview bloğu boş heuristik DEĞİL, assemble_brief → keyword_aio_intel_transform ile gerçek aio.answer_points (+ aio.cited_sources) listesine dönüştürülür ve Brief Paketi'nin aio alanına yazılır. Bu answer_points B2 kapı motorunun aeo kapısı tarafından ölçülür: içerik bu noktaların HER BİRİNİ kapsamalı (aio.present=False ise yalnız "intro cevap-önce" şartı — R-149 / spec §6 R2).

Step 6 — content_gap

cluster_keywords[forbidden_kw] negatif filter (forbidden_reason kontrol). topical_map gap detect: status == 'planned' AND assigned_url == null. Internal link plan: relevant topical_map[pillar/cluster] URL'leri article body içinde reference (per ~300 word 1 link, R-06).

Step 7 — render_5_templates

5 template render-time slot consume (4 file + 1 internal upload section):

  1. templates/content/new-blog.template.md — markdown skeleton (frontmatter + H1/H2/H3 outline). Planning-only, NOT yayınlanır.
  2. templates/content/new-blog.template.html — HTML article fragment (<article class="pse-blog-post"> kök, R-22 fragment boundary).
  3. templates/content/faq-block.template.html — FAQ statik R-43 (accordion <details> + <summary> YASAK, R-43 statik HTML zorunlu).
  4. templates/content/upload-instructions.template.md — Section A new-blog kısmı.

Render-time slot dictionary:

SlotSource
{primary_keyword}new_content_plan.primary_keyword
{tone}brand_identity.tone (Principle 2 profile-aware override)
{hitap}brand_identity.hitap
{primary_color}brand_identity.primary_color
{font_heading}brand_identity.font_heading (alias to font_family_heading)
{font_body}brand_identity.font_body
{author}content_settings.author_strategy switch (YMYL byline-required)
{disclaimer}content_settings.disclaimer_templates[profile_kind]
{default_hero_url}brand_identity.default_hero_url (R-77 fallback)
{image_style}brand_identity.image_style (Principle 2 profile-aware)

CSS class prefix: pse- BEM (plugin agnostik — proje slug hardcode YASAK, e.g. demo-dental/demo-furniture/demo-hvac/demo-petcare/demo-shop/ demo-tires/demo-construction/demo-agency slug'lar HTML/CSS içinde geçemez).

Step 8 — foundational_principles_gate

3-katman gate çalışır (Principle 1 + 2 + 3 yukarıda detaylı). Çıktı: P1 RED → DURUR #6; P2 enum dışı → DURUR #2; P3 7-pattern AMBER 2x → RED upgrade.

Step 9 — render_jsonld_graph (R-78..R-83)

JSON-LD @graph 5 entity render:

  • Article — headline, description, datePublished, dateModified, author (Person ref), publisher (Organization ref), image, mainEntityOfPage.
  • Organization — name, logo (logo_url), sameAs (same_as_urls; R-100).
  • Person — name, jobTitle, knowsAbout (author_strategy switch, R-105 expert quote bank-driven content_settings.experience_database lookup; uydurma yasak — Principle 1 Layer 3).
  • BreadcrumbList — Home > Pillar > Cluster > Article (R-83 4-level cap).
  • FAQPage — talep-güdümlü 3-6 FAQ Q&A items, hard cap 10 (R-79; R-09).

Validate: Google Rich Results Test API mock validate (skill internal — actual API call optional). DURUR #8 trigger: REJECT → schema fix önce.

Step 10 — meta_pixel_wcag_validate

Meta title pixel calc: ≤580px desktop (R-35 single authority; ~60 char is only a tech-audit approximation heuristic — Türkçe glyph widths vary, so the pixel cap is authoritative, not the char count). Meta description pixel calc: ≤990px desktop (R-35; ~160 char heuristic only). DURUR #7 trigger: cap aşımı → auto-correct attempt; fail ise manuel revise.

WCAG 2.1 AA axe-core simulate: violations=0 hedef. Color contrast (primary_color vs background) ≥ 4.5:1. Image alt text required (new_content_plan.alt_text consume; uydurma alt-text yasak — Principle 1).

Step 11 — write_outputs

4 artefact yaz:

  • outputs/blog/{slug}/article.html — <article class="pse-blog-post"> kök, R-22 fragment boundary.
  • outputs/blog/{slug}/schema.jsonld — JSON-LD @graph.
  • outputs/blog/{slug}/meta-tags.json — meta title/description/OG/ Twitter pack.
  • outputs/blog/{slug}/upload-instructions.md — Section A WordPress media library workflow (R-74).

Step 12 — emit_provenance

_state/events.jsonl append (audit):

FieldValue
event_typecontent_new (events.schema event_type enum)
event_kindwork (events.schema enum, ADR-020)
schema_version1.0
actorskill:new-blog
targetmaster.xlsx[new_content_plan]#{id}
timestampUTC ISO 8601
modeinput.mode (draft / publish)

DURUR Conditions (9)

#TriggerResolution
1master.xlsx eksikinit-project önce çalışmalı (Phase 5 W-Q paterni)
2project-config[profile] enum dışı / schema v1.2 yokscripts/migrations/0002_project_config_1.1_to_1.2.py
36 rules dosyası eksikPhase 10 deliverable verify (e4369ea commit)
4MCP budget aşımı (estimated_credits 8)scripts/budget/check_budget.py blocked
5SERP analiz fail (R-08, mode enum dışı)manuel input, awaiting_approval
6P1 fact-check failRED — çıktı discard, claim revize
7meta pixel cap aşımıauto-correct fail ise manuel revise
8Schema Rich Results Test REJECTJSON-LD @graph fix önce
9YMYL profili + reviewer adı yok (R-124 imza eksik)RED yayın bloklu; operatör reviewer adlandır + append_audit kaydı (reviewer+review_date+content_version)

Cascade Fix W-F1 (Phase 10 EKSİĞİ Closure — Atomic Commit Içinde)

Phase 10 finished without project-config[profile] (singular) which Foundational Principle 2 directly cites. Phase 11 W-F1 closes that gap in the SAME atomic commit as the new-blog skill itself, so this skill ships with a schema-valid Principle 2 contract from day one.

File 1 — schemas/project-config.schema.json:

  • schema_version "1.1" → "1.2".
  • properties.profile added (singular, enum 5-value: e-commerce / ymyl / local-service / b2b-saas / portfolio).
  • required[] UNCHANGED (additive policy → rules/schema-versioning-discipline.md).

File 2 — scripts/migrations/0002_project_config_1.1_to_1.2.py:

  • 0001 paterni reuse (idempotent + dry-run + .bak backup).
  • Refuses out-of-range source versions; pure migrate(doc) -> dict function for tests.

File 3 — scripts/state/bootstrap_project.py:

  • SCHEMA_VERSION = "1.1" → "1.2" constant update.

File 4 — tests/skills/test_init_project.py:

  • schema_version assertion "1.1" → "1.2".
  • Asserts schema declares profile field with the exact enum 5-value.

Smoke test (cascade fix verify):

python3 scripts/migrations/0002_project_config_1.1_to_1.2.py \
    --in /tmp/test_config.json --dry-run
jq '.properties.schema_version.const' schemas/project-config.schema.json
jq '.properties.profile.enum'         schemas/project-config.schema.json

Expected: schema_version "1.2", profile.enum array of 5 strings.

Plugin-Agnostic Discipline

  • No project slug hardcoded inside this skill or its templates (demo-dental, demo-furniture, demo-hvac, demo-petcare, demo-shop, demo-tires, demo-construction, demo-agency are FORBIDDEN tokens — pytest grep verifies).
  • CSS class prefix pse- BEM only (pse-blog-post, pse-h2-answer, pse-tldr, pse-faq-block).
  • Brand styling resolves at render-time from brand_identity (per- project), never via hardcoded color/font.
  • Per-call / per-url credit shape names (Phase 7 lesson, ADR-028 anti- pattern) MUST NOT appear in this skill — budget.estimated_credits (number) is the only allowed shape per skill-frontmatter schema.

Error Handling

FailureModeAction
master.xlsx schema driftREDDURUR #1, init-project re-run
project-config[profile] dışıREDDURUR #2, cascade fix verify
MCP DataForSEO budget exhaustedREDDURUR #4, budget gate
Tier-1 Scrapling 403 / blockedAMBERtier-0 SERP description fallback
P1 citation missingREDDURUR #6, claim revize
P3 7-pattern AMBER 2x same passREDmanuel revise
meta pixel cap > 580 / 990 (R-35)AMBERauto-correct → fail RED
Schema Rich Results REJECTREDDURUR #8, @graph fix

References

  • rules/content-quality.md (Foundational Principles, R-01..R-148).
  • rules/content-html-discipline.md (R-22 fragment, R-23 inline CSS, R-43 statik FAQ, R-77 fallback hero).
  • rules/content-seo-discipline.md (R-08 SERP, R-78..R-83 schema).
  • rules/content-eeat-discipline.md (R-28 byline, R-44 source verify, R-104 stats-density single-source, R-105 expert quote, R-114/R-119 research bank, R-124 YMYL pre-publish review sign-off).
  • rules/content-llm-discipline.md (R-109..R-111 AIO citation, R-118 humanize).
  • rules/content-update-discipline.md (R-50 counter-argument).
  • rules/content-craft-discipline.md (R-149..R-153 — CCE pozitif yazım katmanı: pain-mirror giriş, somut örnek, marka sesi, özgünlük, derinlik).
  • scripts/production/new_blog.py (CCE B4 orkestratör yardımcıları — assemble_brief + evaluate; saf Python, MCP/Claude çağırmaz).
  • scripts/production/quality_gates.py (CCE B2 8-kapı motoru — run_all).
  • schemas/project-config.schema.json v1.4 (Phase 3.1 bank entry schema bump — applicable_topics, phrasings, last_used_in_content_id, max_usage_per_month fields R-121 reads; W-F1 v1.2 + Phase 11 cascade still applies for profile singular).
  • schemas/master-excel.schema.json (18 sheets — internal_links and content_gaps are NOT among them; F-4 schema authority).
  • schemas/events.schema.json (event_type enum 12 values, event_kind enum 4 values).
  • schemas/skill-frontmatter.schema.json (8 required fields, category enum 8 values, status enum 3 values).
  • ADR-020 (event_kind=workflow vs work routing).

Wave-Out Note

This skill is status: wip until Phase 11 Wave 2 acceptance gate; active bump deferred to the post-Wave 1 closeout. Manager owns the status flip + governance audit.

Keep looking

Skills are one crate of 613,531. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.