wizz-method 2.11.1 → 2.12.0

package/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "wizz-method",
-  "version": "2.11.1",
+  "version": "2.12.0",
   "description": "Wizz Method: método de agência orientado por IA em PT-BR",
   "keywords": [
     "agile",

package/skills-registry.yaml

@@ -159,7 +159,7 @@ areas:
       - id: wizz-animate
         entry: true
         door: motion
-        when: "PORTA de motion/assets: executar animação e gerar mídia. Craft de micro-interação (regras do Emil Kowalski: saída > entrada, só transform/opacity, reduced-motion). Puxa: wizz-design-motion-principles (auditar animação existente), wizz-remotion-best-practices (vídeo em React), wizz-canvas-design (PNG/PDF por código) e wizz-algorithmic-art (visual generativo)."
+        when: "PORTA de motion/assets: executar animação e gerar mídia. Craft de micro-interação (regras do Emil Kowalski: saída > entrada, só transform/opacity, reduced-motion). Puxa: wizz-design-motion-principles (auditar animação existente), wizz-remotion-best-practices (vídeo em React), wizz-video-effects (efeitos e SFX de ad/Reels com portões), wizz-canvas-design (PNG/PDF por código) e wizz-algorithmic-art (visual generativo)."
         rel:
           pairs_with: [wizz-design-motion-principles]
       - id: wizz-design-motion-principles
@@ -171,7 +171,7 @@ areas:
         door: motion
         when: "MONTAR o filme: vídeo em React (Remotion) com áudio, SFX, legenda e timing com precisão de frame. No pipeline de vídeo é o Remotion que faz a montagem final; o hyperframes só desenha as telas mudas (regra: HyperFrames desenha, Remotion monta — ver _shared/video-pipeline.md do módulo wizz)."
         rel:
-          after: [buttercut, ctc-align, get-brolls, hyperframes, wizz-motion-3d-director]
+          after: [buttercut, ctc-align, get-brolls, hyperframes, wizz-motion-3d-director, wizz-video-effects]
       - id: wizz-canvas-design
         door: motion
         when: "Gerar PNG/PDF editável por código (carrossel, quote card, infográfico) com fonte/cor/logo sempre certos. Traz canvas-fonts."
@@ -187,13 +187,27 @@ areas:
         when: "Timing por forced alignment: timestamps palavra a palavra de narração TTS para sincronizar legenda, corte de cena e animação com a fala. Fonte única de timing no pipeline de vídeo; substitui Whisper e estimativa manual."
         rel:
           after: [voicebox]
+          before: [wizz-remotion-best-practices, wizz-video-effects]
+      # Motor de decisão de efeitos (Python) portado do sistema de efeitos do
+      # trafego em 2026-09-24: orçamento anti-excesso, som por evento semântico,
+      # efeitos nos respiros, portões de zona segura (ad/orgânico), densidade,
+      # excesso, ritmo e loudness. Não monta o filme: isso é do Remotion.
+      - id: wizz-video-effects
+        door: motion
+        when: "Decidir e conferir os EFEITOS e o SOM de um vídeo curto vertical (ad ou Reels): que efeito entra (punch, flash, shake, transição), onde e quando não entra, que SFX toca em cada acontecimento e a que volume, ducking, loudness do master e portões de zona segura, densidade, excesso e ritmo do gancho. Gatilhos: 'efeitos sonoros do criativo', 'SFX do Reels', 'o vídeo tá com zoom demais', 'conferir zona segura do anúncio', 'LUFS do vídeo'."
+        not_when:
+          - "Montar o filme em React (sequência de cenas, legenda, <Audio>): é wizz-remotion-best-practices; esta skill decide os efeitos que ele monta."
+          - "Filme de lançamento de produto com trilha e direção visual: é wizz-huashu-design."
+        rel:
+          after: [ctc-align, hyperframes]
           before: [wizz-remotion-best-practices]
+          pairs_with: [wizz-ad-creative]
     clis:
       - id: hyperframes
         when: "DESENHAR telas de vídeo em HTML/CSS (HTML→MP4, agent-native, 20+ skills): cena isolada e MUDA — maqueta de WhatsApp/ChatGPT, gráfico animado, título, loop. NÃO usar para montagem com áudio/SFX/legenda sincronizada (posicionador de áudio falhou em produção, ~2,76s de drift; CSS vaza entre telas): a montagem final é do Remotion (regra: HyperFrames desenha, Remotion monta — ver _shared/video-pipeline.md do módulo wizz). Precisa ffmpeg + Node 22 + Chrome headless."
         not_when: "Montagem com áudio/SFX/legenda sincronizada (posicionador de áudio falhou em produção, ~2,76s de drift; CSS vaza entre telas)."
         rel:
-          before: [wizz-remotion-best-practices]
+          before: [wizz-remotion-best-practices, wizz-video-effects]
         platform: [darwin, linux]
         # O check confere o SHA fixado, não só a existência da pasta: o bundle é
         # copiado (não clonado), então o pin é gravado num marcador no install.
@@ -283,7 +297,7 @@ areas:
         when: "Timestamps palavra a palavra CONFIÁVEIS de narração TTS (texto conhecido): sincronizar legenda, corte de cena, animação e elemento gráfico com a fala. Fonte única de timing em vídeo — substitui timestamps do Whisper e estimativa manual. Skill da lib: ctc-align (instalação + helper align.py + regras de âncora monotônica e duração mínima). Roda em CPU, sem CUDA."
         rel:
           after: [voicebox]
-          before: [wizz-remotion-best-practices]
+          before: [wizz-remotion-best-practices, wizz-video-effects]
         platform: [darwin, linux]
         check: "test -x ./tools/ctc-align/.venv/bin/ctc-forced-aligner"
         install: 'mkdir -p tools/ctc-align && uv venv --python 3.12 tools/ctc-align/.venv && uv pip install --python tools/ctc-align/.venv/bin/python "git+https://github.com/MahmoudAshraf97/ctc-forced-aligner.git"'
@@ -608,7 +622,7 @@ areas:
       - id: wizz-ad-creative
         when: "Gerar/iterar criativos e headlines de anúncio em escala (RSA, variações)."
         rel:
-          pairs_with: [fal-video, wizz-copywriting, wizz-paid-ads]
+          pairs_with: [fal-video, wizz-copywriting, wizz-paid-ads, wizz-video-effects]
       - id: wizz-analytics-tracking
         when: "Tracking de conversão pros anúncios (compartilhada com growth)."
         rel:

package/src/modules/lowticket/README.md

@@ -43,7 +43,7 @@ The knowledge here is a distillation: facts, benchmarks, checklists and critical
 
 If you extend a shard, keep it that way: write the fact, never the quote. `knowledge/15-como-estender.md` has the full process.
 
-Eight tactics present in the source material are recorded and **not adopted** in `knowledge/guardrails.md`: buying engagement, AI avatars passing as real people, synthetic testimonials, borrowing a third party's ID document to verify a profile, using a real person's name as borrowed proof, reposting someone else's video as your own creative, a struck-through reference price that was never charged, and a hook that asserts a personal attribute of the viewer. They are loaded on every activation and stay blocked until the operation's owner decides otherwise.
+Eight tactics present in the source material are recorded in `knowledge/guardrails.md`, each with a condition of use: buying engagement (risk noted), an AI avatar (with a notice that it is AI), testimonials (real customers only), a third party's document to verify a profile (with the owner's authorization), a real person's name as borrowed proof (with that person's authorization), someone else's video (only as a reference for recording your own), a struck-through price (only if it was actually charged), and a hook with a personal attribute (a writing rule). They are loaded on every activation, and the agent applies each tactic within its condition.
 
 ## Optional config
 

package/src/modules/lowticket/skills/wizz-lowticket-guia/SKILL.md

@@ -37,7 +37,7 @@ Then place them with the symptom table when they describe a problem, or the jour
 | "vende, mas sobra pouco" | average ticket | 7 |
 | "página lenta" / muita gente clica e pouca chega | page speed | detour: `wizz-lowticket-pagina` |
 | "conta bloqueada" / "anúncio reprovado" | account and policy | 6, via `03-campanha-bidcap.md` §3.6 + guardrails |
-| "quero clonar esse funil de quiz" / "achei um quiz validado, quero rodar noutro país" | quiz funnel structural clone | 1, via `wizz-lowticket-minerador` operation E, method in `12b-fabrica-de-ofertas.md` §12.11 - this is the one deliberate exception to "never clone," structure only, needs a coding environment |
+| "quero clonar esse funil de quiz" / "achei um quiz validado, quero rodar noutro país" | mirror the quiz funnel as reference, then rebuild your own | 1, via `wizz-lowticket-minerador` operation E, method in `12b-fabrica-de-ofertas.md` §12.11 - not an exception to "never clone": the mirror is studied, never published, and what goes live is your own funnel; needs a coding environment |
 
 When "gasto e não vendo" comes without numbers, send it to step 6 anyway: `wizz-lowticket-trafego` reads the funnel and will point back to 3, 4 or 5 if that is where the leak is. Do not guess the leak here.
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/SKILL.md

@@ -25,7 +25,7 @@ Never open two shards to find out which one it was. The INDEX has a third column
 
 This matters constantly. `wizz-offer-forge` defaults to a R$10-97 low ticket band and 7/15/30 day guarantees; this base pins price by source band (R$19,90 / R$27 / R$37), upsell at 2 to 2.5x the entry, and has an unresolved point about 7 vs 14 day guarantees. Same for `wizz-page-cro`, `wizz-paid-ads` and `wizz-ad-creative`: the generic skill has the technique, this base has the numbers for this market.
 
-The one thing that outranks a shard is `knowledge/guardrails.md`. Eight tactics present in the source material are recorded there and **not adopted**: buying engagement, AI avatars passing as real people, synthetic testimonials, borrowing a third party's ID document to verify a profile, using a real person's name as borrowed proof, reposting someone else's video as your own creative, a struck-through reference price that was never charged, and a hook that asserts a personal attribute of the viewer. They are blocked until the operation's owner decides otherwise. If a shard teaches one, the inline block line is there too.
+The one thing that outranks a shard is `knowledge/guardrails.md`. Eight tactics present in the source material are recorded there, each **with a condition of use** (the operation owner's decision, 2026-09-24): buying engagement, allowed with the risk noted; an AI avatar, with a notice that it is AI; testimonials, only from real customers; a third party's document to verify a profile, with the document owner's authorization; a real person's name as borrowed proof, with that person's authorization; someone else's TikTok video, only as a reference for recording your own; a struck-through price, only if it was actually charged (otherwise anchor by value without striking through); and a hook with a personal attribute, under the writing rule (occupation, role, result, or goal pass; a sensitive trait in the second person fails). Apply each one with its condition from `guardrails.md`; when the condition isn't met, use the alternative in the same section. If a shard teaches one, the pointer to the condition is inline there too.
 
 ## Flow
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/02-diagnostico-funil.md

@@ -11,19 +11,21 @@ Funnel base = Page Visits = 100%.
 
 | # | Transition | Bad | Average | Healthy |
 |---|-----------|------|-------|----------|
-| 1 | Ad Clicks → Page Visits | < 70% | 70–90% | ≥ 95% |
-| 2 | Page Visits → Initiate Checkout (IC) | < 10% | 10–25% | 25–40%+ |
+| 1 | Ad Clicks → Page Visits (connect rate) | < 85% (urgent) | 85–90% (optimize) | > 90% |
+| 2 | Page Visits → Initiate Checkout (IC) | < 10% (redo the page) | 10–25% (optimize) | 25–40%+ |
 | 3 | Initiate Checkout → Sale Initiated | < 50% | 50–75% | 75–90% |
 | 4 | Sale Initiated → Sale Approved (PIX) | < 50% | 60–75% | 75–90% |
 | 5 | Sale Initiated → Sale Approved (Card) | < 60% | 60–75% | 75–90% |
 
+**Connect rate (transition 1):** always measure on mobile with **4G throttling**. Single ruler adopted on 2026-09-24 ([point 62](pontos-em-aberto.md)): above 90% is good, between 85% and 90% means optimize, below 85% is urgent. **Visits→IC (transition 2):** below 10% redo the page; between 10% and 20% optimize it, don't discard it ([point 61](pontos-em-aberto.md)).
+
 **Calculation:** the real rate for each transition is one step divided by the previous one, not by the fixed base. E.g.: `IC / Visits × 100`.
 
-### 2.2 Transition 1: Clicks → Visits (< 95% = a technical problem, not a marketing one)
+### 2.2 Transition 1: Clicks → Visits (< 90% = a technical problem, not a marketing one; < 85% = urgent)
 
-**Causes:** slow LP (>3s on 3G), broken or looping redirect, SSL error, LP broken on mobile (95% of traffic is mobile), Pixel not firing `PageView` (here the measurement is wrong, not the conversion).
+**Causes:** slow LP (>3s on 4G), broken or looping redirect, SSL error, LP broken on mobile (95% of traffic is mobile), Pixel not firing `PageView` (here the measurement is wrong, not the conversion).
 
-**Actions:** test the LP with 3G throttling in DevTools; run PageSpeed Insights and Core Web Vitals (LCP, CLS, INP); click straight to the LP without a link tracker in between; confirm the Pixel and `PageView`.
+**Actions:** test the LP with 4G throttling in DevTools; run PageSpeed Insights and Core Web Vitals (LCP, CLS, INP); click straight to the LP without a link tracker in between; confirm the Pixel and `PageView`.
 
 > If the leak is here, it's dev work. Fix it and come back. Don't touch the creative.
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/03-campanha-bidcap.md

@@ -15,18 +15,18 @@
 
 ### 3.2 Bid and budget rules
 
-- Don't lower the BidCap while performance is high. With RPU of R$20–23, keep the bid and give Facebook room to spend.
+- **Don't lower the BidCap while performance is high** (this also holds after 3 green days; [point 57](pontos-em-aberto.md)). With RPU of R$20–23, keep the bid and give Facebook room to spend.
 - **Max CPA = break-even + 30%.** That's the acceptable margin of error.
 - Test larger budgets (e.g. R$500) within that margin.
 - **10-day window:** Facebook takes up to 10 days to regain intelligence after changes. Don't keep testing indefinitely beyond that.
-- *Source for the four lines above and the 3.3 cutoff: an auto-generated transcript, unverified. The numbers (RPU R$20-23, +30%, R$500, 10 days, ROI 1.3) come from that transcript, not checked against the recording.*
+- *Source for the four lines above and the former ROI 1.3 cutoff in 3.3: an auto-generated transcript, unverified. The numbers (RPU R$20-23, +30%, R$500, 10 days, ROI 1.3) come from that transcript, not checked against the recording.*
 - **Vertical scale by ROAS** (adopted rule, [open point 54](pontos-em-aberto.md)): **ROAS ≥ 2.5 → +50%** · **ROAS between 2.0 and 2.5 → +30%** · **ROAS < 2.0 → +20%**. Replaces the flat +30%-a-day ceiling, which now only applies to the middle band.
-- **Initial bid cap anchored to the average ticket, with slack:** the source material says to start around 50% above the ticket and adjust daily as sales come in. E.g.: R$50 ticket → bid cap around R$75. The auto-generated transcript contradicts itself here: the table says "starting at 50% of the average ticket," the literal quote says "50% more on the Bid Cap"; the quote wins.
+- **Initial bid cap = break-even × 1.5** (rule adopted on 2026-09-24, [point 58](pontos-em-aberto.md); calculation in 3.7), adjusted daily as sales come in. E.g.: R$40 break-even → initial bid cap of R$60. One source anchored the +50% to the average ticket; with thin margins that gives a different number, and break-even wins.
 
 ### 3.3 Analysis and cuts
 
 - **The unit of analysis is the individual ad**, not the ad set.
-- **Turn off a creative with ROI < 1.3.** Keep the winners active.
+- **Only turn off a creative when all 4 conditions of 3.7 line up:** spend ≥ 3x the Bid Cap, ROAS below break-even, zero sales and no recent structural change (rule adopted on 2026-09-24, [point 59](pontos-em-aberto.md); the transcript said ROI < 1.3). Keep the winners active.
 - This lets you test new creatives without affecting the winner.
 
 ### 3.4 When to scale
@@ -63,7 +63,7 @@ The source material treats low CPM/CPC as a result of **account infrastructure**
 
 **Account age isn't the factor.** A T1 account versus a T4 account makes no meaningful difference for the initial test. What actually makes an account "fall" in practice is a payment or verification error, not account age. The account has to stay active and monitored.
 
-> ⚠️ **Buying engagement — recorded, not recommended.** The source material mentions a paid tool for buying likes and comments on videos to inflate engagement signals for the algorithm. This is **inauthentic engagement and violates Meta's Terms**, with the risk of the account and the Business Manager being banned — exactly the assets the rest of this section is trying to protect. It's recorded here because it's in the source material, but it isn't adopted as a recommended practice. See [guardrails 20](guardrails.md).
+> ⚠️ **Buying engagement: allowed, with the risk noted.** The source material mentions a paid tool for buying likes and comments on videos to inflate engagement signals for the algorithm. Under Meta's Terms it is inauthentic engagement: Meta may cut reach or take down the account and the Business Manager, exactly the assets the rest of this section protects. Condition and risk in [guardrails 20](guardrails.md).
 
 ### 3.7 The traffic manager's ruler
 
@@ -73,9 +73,8 @@ This file is the source's **operational campaign-diagnosis prompt**. The rulers
 
 **Bid Cap calculation:**
 
-- Base: **Bid Cap = Breakeven**.
-- Aggressive: Breakeven × 1.30 or × 1.50.
-- **Current operational recommendation: +50% over breakeven.**
+- **Initial Bid Cap = Breakeven × 1.5** (the prompt's operational recommendation, adopted as the single ruler on 2026-09-24, [point 58](pontos-em-aberto.md)).
+- The prompt also lists Bid Cap = Breakeven (base) and Breakeven × 1.30 as variants; they are not the starting point.
 - Prerequisites for using Bid Cap: breakeven calculated, at least 1 validated creative, Pixel ok, CAPI ok, and enough data. Without that, start with **Highest Volume until ~20-30 purchases**.
 
 **The 3x ruler — the most important one in this section:** don't judge, don't pause and don't change anything before the new campaign has accumulated **spend of 3x the Bid Cap**. Below 24h is validation, not a result.
@@ -86,7 +85,7 @@ This file is the source's **operational campaign-diagnosis prompt**. The rulers
 - **Hold Rate** = 75% plays ÷ impressions.
 - ROAS: use the "Results ROAS" column directly, **never** calculate value ÷ cost by hand.
 
-**Only pause a creative when all 4 line up:** spend ≥ 3x Bid Cap **and** ROAS below breakeven **and** zero sales **and** no recent structural change.
+**Only pause a creative when all 4 line up** (the single cutoff ruler, [point 59](pontos-em-aberto.md)): spend ≥ 3x Bid Cap **and** ROAS below breakeven **and** zero sales **and** no recent structural change.
 
 **Image CTR isn't comparable to video CTR:** image CTR is naturally higher. Decide on post-click conversion instead.
 
@@ -117,7 +116,7 @@ This file is the source's **operational campaign-diagnosis prompt**. The rulers
 **Portugal · validate before scaling:**
 
 - **Validation isn't the first sale.** The offer only counts as validated with **3 straight days of positive ROI and stable metrics**, in the range of **30 to 50 sales on the Pixel**. Log daily CPC, CTR and ROAS from day one, to have a comparison baseline.
-- **Starting budget** on CBO Highest Volume: **2x to 3x the average ticket per day**, with a bid around **1.5x the ticket** (the same +50% slack over the ticket from 3.2; the 3.7 ruler applies the +50% over **breakeven**, not over the ticket).
+- **Starting budget** on CBO Highest Volume: **2x to 3x the average ticket per day**, with an initial bid of **break-even × 1.5** (the ruler of 3.2 and 3.7; the source said 1.5x the ticket, [point 58](pontos-em-aberto.md)).
 - **A new creative goes into a new ad set** inside the same campaign, never duplicating the original set: it holds the learning history. CBO Highest Volume finds the Top Spenders; Bid Cap then scales only those.
 - **One account only until validated.** A small market can't support several accounts on the same not-yet-validated offer; extra accounts (CBO or Bid Cap) only after validation.
 - **Vertical scale conditioned on ROAS:** +50% with ROAS ≥ 2.5, +20% with ROAS < 2.0 (the source doesn't mention the 2.0-2.5 band, which landed at +30%). Became the general 3.2 ruler ([open point 54](pontos-em-aberto.md)): **ROAS ≥ 2.5 → +50% · 2.0-2.5 → +30% · < 2.0 → +20%**.

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/04-criativos.md

@@ -8,7 +8,7 @@
 
 ### 4.1 Creative structure
 
-1. **Hook** (first 3 seconds): captures and filters. *(11.6 puts it at 5 seconds; see [open point 19](pontos-em-aberto.md).)*
+1. **Hook** (first 3 seconds): captures and filters. *(adopted rule, see [open point 19](pontos-em-aberto.md).)*
 2. **Body**: useful content, social validation, an engaging story.
 3. **CTA**: urgent call to action.
 
@@ -16,7 +16,7 @@
 
 - **Minimum of 10 creatives** with distinct angles and formats to validate. Never depend on a single "hulk" creative.
 - Structured setup: **5 different angles per ad set**; duplicate the 5 originals per variation (e.g.: 5 with a female voice, 5 with a male voice).
-- **Ideal length: 40 to 60 seconds.** Facebook's 15s recommendation is ineffective. Max 90s, above that there's a risk of being blocked. *(Against the checklist's 90s-2min VSL and the VSL benchmark: see [open point 11](pontos-em-aberto.md). Against the 15-30s of §4.18 and the 3-5min remarketing of §4.12 item 4: see [open points 50](pontos-em-aberto.md) and [60](pontos-em-aberto.md).)*
+- **Ideal length: 40 to 60 seconds.** Facebook's 15s recommendation is ineffective. On **cold traffic, max 90s**, above that there's a risk of being blocked. **Exception: remarketing goes up to 5 min** (§4.12 item 4); the split by length is in [`04c`](04c-criativos-roteiro.md) §4.18 (rule adopted on 2026-09-24, [point 60](pontos-em-aberto.md)). *(Against the checklist's 90s-2min VSL and the VSL benchmark: see [open point 11](pontos-em-aberto.md). Against the 15-30s of §4.18: see [open point 50](pontos-em-aberto.md).)*
 - 9:16 format, dynamic cuts, no pauses.
 - Launch new creatives **every week**.
 - Replace image creatives with video (VSL) to raise conversion.
@@ -42,7 +42,7 @@ Instagram users are looking for entertainment, not a sale. The creative has to l
 
 **Minimum anatomy:** immediate visual hook ("hulk") → body showing the product → CTA.
 
-- **Name the audience in the first second.** "Você que é lojista..." (you, the shop owner...) creates immediate identification and filters out who isn't. **Occupation or role only:** a sensitive trait stated in the second person ("você que está endividado") is blocked — [guardrails 49](guardrails.md).
+- **Name the audience in the first second.** "Você que é lojista..." (you, the shop owner...) creates immediate identification and filters out who isn't. **Occupation or role only:** also result or goal ("pra quem quer sair das dívidas"); a sensitive trait stated in the second person ("você que está endividado") fails, writing rule in [guardrails 49](guardrails.md).
 - **The 80% image rule.** Plenty of images, very little text. Show the product, don't describe the product.
 - **Showing the exact quantity** ("80 figurinhas" - 80 stickers) creates more desire than showing a sample.
 - **Organic format beats produced format.** Simple static images and video with dynamic cuts beat heavy editing.

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/04a-criativos-ia.md

@@ -5,11 +5,11 @@
 
 **What's here:** `4.6` realism in AI video · `4.9` realistic images and avatars · `4.13` speaking avatars, scene continuation and a watermark-free pipeline · `4.14` 20-minute pipeline for TikTok · `4.15` bank of free tools
 **Siblings:** fundamentals in [`04-criativos.md`](04-criativos.md) · script, templates and swipe file in [`04b-criativos-formatos.md`](04b-criativos-formatos.md) · writing the video script in [`04c-criativos-roteiro.md`](04c-criativos-roteiro.md)
-**Attention:** an AI avatar passing as a real person is a blocked tactic — see [guardrails 21](guardrails.md).
+**Attention:** an AI avatar is allowed with a notice that it is AI; see the condition in [guardrails 21](guardrails.md).
 
 ### 4.6 Realism in AI video
 
-> ⚠️ Realistic AI avatar: see [guardrails 21](guardrails.md) before presenting a generated avatar as if it were a real person in an ad.
+> ⚠️ Realistic AI avatar: it goes with a notice that it is AI, see the condition in [guardrails 21](guardrails.md).
 
 Two tools, distinct functions: **Dream Face** for lip-sync (a speaking avatar) and **Eleven Labs** for voice generation.
 
@@ -29,7 +29,7 @@ Two tools, distinct functions: **Dream Face** for lip-sync (a speaking avatar) a
 
 ### 4.9 Realistic images and avatars with AI — Google Flow
 
-> ⚠️ **AI avatar as a real person — confirm before using.** Meta and Instagram require an account representing a fictitious person to be identifiable as such, and the rule varies by format and by country (stricter in Portugal and the EU, see the EU compliance note in §4.8 of [`04b`](04b-criativos-formatos.md)). The source material doesn't address this. An AI avatar as a declared visual device, yes; an AI avatar presented as a real person in an ad, **not without confirming the target market's policy**. See [guardrails 21](guardrails.md).
+> ⚠️ **AI avatar: allowed with a notice that it is AI.** Meta and Instagram require an account representing a fictitious person to be identifiable as such, and the rule varies by format and by country (stricter in Portugal and the EU, see the EU compliance note in §4.8 of [`04b`](04b-criativos-formatos.md)). The avatar itself is method; what you don't do is hide that it's AI. Condition in [guardrails 21](guardrails.md).
 
 Complements 4.6 (which covers spoken video with Dream Face + Eleven Labs). Here the object is **generating the photo/avatar**, for a profile or for a creative.
 
@@ -80,7 +80,7 @@ Generate a realistic vertical selfie video format. The character should speak na
 
 ### 4.14 20-minute pipeline — video creative for TikTok
 
-> ⚠️ AI avatar presented as a real person: see [guardrails 21](guardrails.md). The lip-sync rule also applies: never reuse an avatar take whose mouth doesn't match the piece's audio 1:1.
+> ⚠️ An AI avatar goes with a notice that it is AI: see the condition in [guardrails 21](guardrails.md). The lip-sync rule also applies: never reuse an avatar take whose mouth doesn't match the piece's audio 1:1.
 
 Same stack as §4.6, §4.9 and §4.13 — Google Flow, Eleven Labs, CapCut. What changes here is the **order of operations and the timing**, not the tooling.
 
@@ -105,7 +105,7 @@ Same stack as §4.6, §4.9 and §4.13 — Google Flow, Eleven Labs, CapCut. What
 
 ### 4.15 Tool bank — free alternatives
 
-> ⚠️ **Two tactics from this lesson are not adopted.** Downloading someone else's TikTok video (with or without watermark) to run as your own creative, and using a photo of a person from another country to "test acceptance" without usage rights. Reason and alternative in [guardrails 45](guardrails.md). The rest of the lesson — free stock banks, a voice alternative, TikTok as a reference to STUDY format — still stands.
+> ⚠️ **Someone else's TikTok video: only as a reference for recording your own.** Model the video that went viral (format, pace, hook) and record your own; don't upload the other person's video, with or without a watermark, and don't use a photo of a person from another country without usage rights. Condition in [guardrails 45](guardrails.md). The rest of the lesson (free stock banks, a voice alternative, TikTok as a reference to study format) still stands.
 
 The part of TikTok that remains valid is using the platform to **observe and model format** (what's working, how they're cutting it, what hooks they use), never to extract someone else's file and reuse it as your own.
 
@@ -118,7 +118,7 @@ The part of TikTok that remains valid is using the platform to **observe and mod
 | **Magnific** (transcript spells it "Magnifique") | Image generation/processing | — |
 | **Pexels** | Free image and video stock bank | — |
 | **TikTok** | Source of B-roll and reference for studying format | Only for observing/modeling — see warning above, never for downloading and reusing someone else's video |
-| "Coque Coque" in the transcript — probably the **Cốc Cốc** browser, which has a built-in video downloader | Downloading TikTok video without a watermark | **Name not confirmed: verify before installing.** And the use itself is blocked — see warning above |
+| "Coque Coque" in the transcript — probably the **Cốc Cốc** browser, which has a built-in video downloader | Downloading TikTok video without a watermark | **Name not confirmed: verify before installing.** Download only to study as a reference; the other person's video is not uploaded (see warning above) |
 
 **Not identified:** the transcript mentions "Eddes" in comparison with Eleven Labs, without enough context to know which tool it refers to. Recorded here as pending, not as a recommendation.
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/04b-criativos-formatos.md

@@ -133,7 +133,7 @@ Material cited in a lesson. **This is the creative's visual-framing taxonomy**,
 | **LEIA A LEGENDA** | curiosity headline that pushes to the caption | "Use esse como headline gerando curiosidade" (use this as a curiosity-generating headline) |
 | **OFERTA DIRETA** | struck-through price + final price + CTA | "SUA PROMESSA MAIS FORTE · Por apenas: 0,00" (your strongest promise · for only: 0.00) |
 
-> ⚠️ **`OFERTA DIRETA` with a struck-through price:** only strike through a price that was actually charged before. Striking through the sum of the value assigned to the materials is a blocked tactic, see [guardrails 48](guardrails.md); the alternative is declared value anchoring ("€134 in materials · today €14.90"), with no strikethrough.
+> ⚠️ **`OFERTA DIRETA` with a struck-through price:** only strike through a price that was actually charged before. If it was never charged, it doesn't get struck through (condition in [guardrails 48](guardrails.md)); the alternative is declared value anchoring ("€134 in materials · today €14.90"), with no strikethrough.
 
 **The principle that ties almost all of them together: the creative disguises itself as native content.** Nine of the 14 templates mimic an interface the user already knows and trusts — a gossip post, a news item, a tweet, a livestream, a video call, an Instagram sticker. It's the literal execution of rule 4.4: *"the creative has to look native to the feed and disguise the commercial pitch inside content of value."* The material itself says so on the news-format page: **"disguising your ad using the news format converts very well with the 50+ audience."**
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/04c-criativos-roteiro.md

@@ -5,7 +5,7 @@
 
 **What's here:** `4.16` the minimum brief · `4.17` the four decisions before the first line · `4.18` block map by duration · `4.19` rules per block · `4.20` delivery format · `4.21` testing one axis per batch · `4.22` retention diagnosis · `4.23` pre-delivery checklist · `4.24` modular production: hook × body × CTA
 **Siblings:** fundamentals in [`04-criativos.md`](04-criativos.md) · AI production in [`04a-criativos-ia.md`](04a-criativos-ia.md) · VSL, templates and swipe file in [`04b-criativos-formatos.md`](04b-criativos-formatos.md)
-**Attention:** a hook that asserts a personal attribute of the viewer is blocked — see [guardrails 49](guardrails.md).
+**Attention:** a hook with a personal attribute follows the writing rule in [guardrails 49](guardrails.md): occupation, role, result or goal pass; a sensitive trait asserted in the second person fails.
 
 **Why this shard exists.** Section 4 says what a creative is made of and [`18`](18-copy-desejo-oferta.md) says why a sentence sells. Neither says how to turn both into a script someone can **record today**. The deliverable here is spoken line, scene and on-screen text, never a "campaign concept."
 
@@ -139,10 +139,10 @@ Complements §18.12, which starts at the CTR. This table reads **inside** the vi
 - [ ] At least one real proof, cited by its origin
 - [ ] One CTA, literal, repeated
 - [ ] Every sentence fits in ~12 words and passes the read-aloud test
-- [ ] No personal attribute stated about the viewer ([guardrails 49](guardrails.md))
+- [ ] No sensitive trait asserted about the viewer in the second person ([guardrails 49](guardrails.md))
 - [ ] No guaranteed result, cure or certain gain: use a range, a timeframe and the source of the number
 - [ ] No invented study, institution, specialist, testimonial or print ([guardrails 23](guardrails.md))
-- [ ] No real person's name or image implying endorsement ([guardrails 44](guardrails.md))
+- [ ] No real person's name or image without their authorization ([guardrails 44](guardrails.md))
 - [ ] Table **and** clean spoken text delivered
 
 When a request crosses one of these lines, deliver the version that works **and** is true, and say in one sentence what changed and why.

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/06-pagina-vendas.md

@@ -137,7 +137,7 @@ This creates **two clocks**, and both need optimizing:
 - In audits, **traffic** campaigns optimized for page views showed numbers far above the real sessions measured in analytics. In **conversion** campaigns, the numbers matched.
 - Practical consequence: run a purchase-optimized campaign and create a **custom conversion based on the page URL** to measure connect rate without guessing. Cross-check with Utmify before deciding anything, because the ads manager rounds up.
 
-> **A connect rate below 85% is almost never a bad audience. It's a bad road.** And if the connect rate is already above 92%, the problem is somewhere else: speed doesn't fix a bad offer, a bad creative, or a bad checkout.
+> **A connect rate below 85% is almost never a bad audience. It's a bad road.** And if the connect rate is already above 90%, the problem is somewhere else: speed doesn't fix a bad offer, a bad creative, or a bad checkout.
 
 ### 6.4 Technical performance targets for the page
 
@@ -176,6 +176,6 @@ This creates **two distinct clocks** to optimize: the time until the pixel fires
 
 > Speed fixes the leak between the click and the page, and only that. It doesn't fix a bad offer, a bad creative, or a bad checkout. With a connect rate above 92%, the problem is somewhere else.
 
-Above 92%, stop optimizing the page and go back to the hierarchy in section 2.6.
+Above **90%**, stop optimizing the page and go back to the hierarchy in section 2.6; between 85% and 90%, optimize; below 85%, it's urgent. *(Single ruler adopted on 2026-09-24, [point 62](pontos-em-aberto.md), the same as [`02`](02-diagnostico-funil.md) §2.1; the quote above says 92%.)*
 
 **Player solution cited:** a custom video player built on top of YouTube, with controls similar to VTurb's, to simulate the full purchase session.

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/06a-instagram.md

@@ -3,7 +3,7 @@
 > Shard of this base's methodology. Index: [`INDEX.md`](INDEX.md) · Always loaded alongside: [`guardrails.md`](guardrails.md)
 > Where the generic skill contradicts this shard, **the shard wins**, and this must be stated in the response.
 
-Source: + `mapa-mental-instagram-forte` (the lesson's site, 2026-08-17). Complements and details section 11.2 of the ROI checklist.
+Source: + `mapa-mental-instagram-forte` (the lesson's site, 2026-08-17). Complements and details section 11.2 of the ROI checklist, which since 2026-09-24 follows the profile anatomy here ([point 63](pontos-em-aberto.md)).
 
 ### 6A.1 The proof: where the sales come from
 
@@ -57,7 +57,7 @@ Pass all five, they buy through the bio or the story. Fail one, they leave — a
 - **Reported effect:** lower CPA and CPM even without the badge. **Meta doesn't promise this.** Meta Verified and Meta One offer the badge, impersonation protection and support, nothing about ad cost. Treat it as a hypothesis to test, never as a reason to verify.
 - Especially useful for operating in markets like Portugal.
 
-> The source also teaches submitting a **third party's** document (the example given is using the mother's document) to pass verification. **This project does not do that.** See [`guardrails.md`](guardrails.md) item 32 and [Open item 32](pontos-em-aberto.md). Under Instagram's rules, a badge obtained with false or misleading information is removed and the account may be disabled. The risk lands on the profile that carries the offer.
+> The source also teaches submitting a **third party's** document (the example given is using the mother's document) to pass verification. **Allowed with the document owner's authorization**: condition in [`guardrails.md`](guardrails.md) item 32 and [Open item 32](pontos-em-aberto.md). Under Instagram's rules, a badge obtained with false or misleading information is removed and the account may be disabled. The risk lands on the profile that carries the offer.
 
 ### 6A.4 Highlights — organize by objection, not by topic
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/07-funil-upsell.md

@@ -27,7 +27,7 @@
 
 PIX + card + WhatsApp with a unique code for immediate recovery. WhatsApp drastically reduced pending payments, more than simply disabling boleto (bank slip). An indebted audience prefers PIX, and boleto can pollute the campaign's intelligence.
 
-**Portugal exception (MB WAY).** MB WAY (often misspelled "MDWay" in raw source notes) is the local Pix equivalent on Hotmart, and **it blocks the order bump**: offering the bump on it triggers an objection instead of a sale, so the bump comes off the entire checkout for whoever pays by MB WAY (this is a one-off exception to the "order bump only on the premium plan" rule in 5.3, not a replacement for it — it's a payment-platform limitation, not a copy one). Compensate on the upsell instead: a price of **14€ to 29€** *(figure from an auto-generated summary, unverified; the source's literal transcript puts the front at up to €19 and U1 at €29 "to start", and the range clashes with 7.4's 40-60% of the front: not settled)* (not the "cheap front" that works in Brazil), enough to cover Hotmart's VAT pass-through to the end customer and still keep margin per customer — necessary because the Portuguese market only has ~8 million reachable people, and creative scale doesn't offset a low ticket the way it does in Brazil. A personalized thank-you page with a direct checkout button (instead of one-click) avoids a payment block from the local method.
+**Portugal exception (MB WAY).** MB WAY (often misspelled "MDWay" in raw source notes) is the local Pix equivalent on Hotmart, and **it blocks the order bump**: offering the bump on it triggers an objection instead of a sale, so the bump comes off the entire checkout for whoever pays by MB WAY (this is a one-off exception to the "order bump only on the premium plan" rule in 5.3, not a replacement for it — it's a payment-platform limitation, not a copy one). Compensate on the upsell instead: **front up to €19 and U1 at €29** *(the source's literal words, "to start"; rule adopted on 2026-09-24, [point 64](pontos-em-aberto.md); an auto-generated summary of the same source said 14€ to 29€)* (not the "cheap front" that works in Brazil), enough to cover Hotmart's VAT pass-through to the end customer and still keep margin per customer — necessary because the Portuguese market only has ~8 million reachable people, and creative scale doesn't offset a low ticket the way it does in Brazil. A personalized thank-you page with a direct checkout button (instead of one-click) avoids a payment block from the local method.
 
 ### 7.4 The funnel's price and conversion ruler
 
@@ -37,7 +37,7 @@ Source operational prompt that turns a main product into a full post-checkout fu
 
 | Step | Price |
 |---|---|
-| **U1** | 40% to 60% of the front price |
+| **U1** | 40% to 60% of the front price *(Portugal is the exception: U1 at €29 with a front up to €19, see 7.3 and [point 64](pontos-em-aberto.md))* |
 | **U2** | same as U1, or up to ~20% above |
 | **Downsell** | 10% to 25% of the price of the product that was **declined** |
 | **Final Offer** | a practical floor of ~R$9 to R$15 in the Brazilian market |

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/11-checklist-roi.md

@@ -12,12 +12,12 @@
 
 ### 11.2 The offer's Instagram (BBF)
 
-- A profile dedicated only to the offer, never the personal one. Bio in 2 lines: numeric proof + promise + arrow to the link. 3 highlights: deliverable on screen, customer feedback, FAQ. **A sales story every day + a seal** — drives down CPC and raises Pix approval (practitioner report; Meta makes no claim about the badge lowering ad cost, see [06a §6A.3-A](06a-instagram.md)).
+- A profile dedicated only to the offer, never the personal one. Bio in 4 lines (who, proof, who it's for, CTA) and 5 to 7 highlights organized by objection, with deliverable on screen, customer feedback and FAQ among them; full anatomy in [06a §6A.3-6A.4](06a-instagram.md) (rule adopted on 2026-09-24, [point 63](pontos-em-aberto.md)). **A sales story every day + a seal** — drives down CPC and raises Pix approval (practitioner report; Meta makes no claim about the badge lowering ad cost, see [06a §6A.3-A](06a-instagram.md)).
 
 ### 11.3 VSL and page
 
 - Low-ticket VSL: **90 seconds to 2 minutes** *(when the product proves itself on screen; an offer that needs to build a mechanism and authority goes to 7-9 min, see [Open item 18](pontos-em-aberto.md))* (it can't sustain a 20-min VSL). Content: screen recording opening up the product from the inside (**showrun**). No take longer than 2.5s; the soundtrack follows the emotion. Final CTA teaches how to pay: where to click, what to fill in, when access lands.
-- Page with **two plans** (basic = lead magnet, premium = the sale) + a pop-up on the basic click offering the premium for less. *(Rule adopted on 2026-08-28: from R$47 up. Up to R$37 it's a single price, per the anatomy in 12.7-A. See [Open item 25](pontos-em-aberto.md).)* Lightweight page: **connect rate > 90%**. **Page conversion > 25%; below 20% means redo the page, not the traffic.**
+- Page with **two plans** (basic = lead magnet, premium = the sale) + a pop-up on the basic click offering the premium for less. *(Rule adopted on 2026-08-28: from R$47 up. Up to R$37 it's a single price, per the anatomy in 12.7-A. See [Open item 25](pontos-em-aberto.md).)* Lightweight page: **connect rate > 90%** (measured on mobile with 4G throttling; 85-90% means optimize, below 85% is urgent; [point 62](pontos-em-aberto.md)). **Page conversion (Visits→IC) > 25% is the target; between 10% and 20% means optimize the page; below 10% means redo the page, not the traffic** ([point 61](pontos-em-aberto.md)).
 
 ### 11.4 Checkout and margin
 
@@ -30,7 +30,7 @@
 
 ### 11.6 Creative and campaign
 
-- Hook in the first 5 seconds, captions always on, 9:16, no studio look. **Link CTR > 1.5% and CPC < R$2** — outside that range the problem is the creative.
+- Hook in the first 3 seconds (see [open point 19](pontos-em-aberto.md)), captions always on, 9:16, no studio look. **Link CTR > 1.5% and CPC < R$2** — outside that range the problem is the creative.
 - **Rule for reading the two metrics**: CTR below 1.5% = creative (weak hook, nobody clicks). CPC above expected with good CTR = audience or VSL, not creative. While either metric is off, **don't touch the page, order bump, or upsell** — the leak is upstream. Broad audience: **the creative does the segmenting, not the interest targeting**. Found the winner: new variations and angles **within the same campaign**.
-- Campaign: CBO with Advantage on everything (per the source, ABO is going away). **Bid cap at the full average ticket** (never half). Inflated budget: a R$30 ticket calls for a R$300-500/day campaign. Zero targeting (age and placement on automatic).
-- Verdict rules: spent almost nothing in the first **17h** = the offer didn't pass (tested for free). Spent **3x the ticket with ROI < 1** = kill it (don't touch it before that). **3 days of positive ROI** = lower the bid cap and raise the budget by at most +50%. *(The step size follows the ROAS ruler from §3.2 of [`03`](03-campanha-bidcap.md), adopted 2026-09-24 ([open point 54](pontos-em-aberto.md)): +50% only with ROAS ≥ 2.5.)*
+- Campaign: CBO with Advantage on everything (per the source, ABO is going away). **Initial bid cap = break-even × 1.5** (§3.2 and §3.7 of [`03`](03-campanha-bidcap.md), [point 58](pontos-em-aberto.md); the source said "full average ticket, never half"). Inflated budget: a R$30 ticket calls for a R$300-500/day campaign. Zero targeting (age and placement on automatic).
+- Verdict rules: spent almost nothing in the first **17h** = the offer didn't pass (tested for free). A creative is only cut with **all 4 conditions of §3.7** of [`03`](03-campanha-bidcap.md): spend ≥ 3x the Bid Cap, ROAS below break-even, zero sales and no recent structural change (don't touch it before that; [point 59](pontos-em-aberto.md)). **3 days of positive ROI** = offer validated: **don't lower the bid cap** while performance is high (§3.2, [point 57](pontos-em-aberto.md)); scale in the lever order of §3.7 (raise the Bid Cap first) with the step from the ROAS ruler of §3.2 (≥ 2.5 → +50% · 2.0 to 2.5 → +30% · < 2.0 → +20%; [point 54](pontos-em-aberto.md)).

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/12-mineracao-ofertas.md

@@ -47,7 +47,7 @@ Justify always broken down by criterion (e.g., "3+2+1+2+1 = 9"). Score <6 doesn'
 
 ### 12.3 From data to MVP and the right order
 
-The spreadsheet row gives birth to the MVP: name and positioning, deliverable in 4 PDFs + 1 bonus unlocked after 7 days (reduces refunds), section-by-section page structure, 5 creative angles in image form. (The 5: curiosity, concrete pain, before/after, authority, urgency, one per image, 1:1 format in the source. Rule adopted on 2026-09-04: **4:5** format and 5 angles chosen from 7, see [`12c`](12c-anatomia-low-ticket.md) §12.7-A and [points 35](pontos-em-aberto.md) and [36](pontos-em-aberto.md).) **The right order:** the spreadsheet delivers the offer → MVP in minutes → test with a small budget → if it doesn't sell, AI analyzes the numbers → if it sells, **raise the budget 30% at a time, never double it** ("doubling resets the learning"). Current vertical-scale ruler since 2026-09-24: the ROAS-based one in [`03`](03-campanha-bidcap.md) §3.2 ([point 54](pontos-em-aberto.md)); the source's 30% only applies for ROAS between 2.0 and 2.5 (diverges from 03 §3.2's ROAS-based ruler, point 54; whether it still applies at this freshly-validated MVP stage is [open point 56](pontos-em-aberto.md)).
+The spreadsheet row gives birth to the MVP: name and positioning, deliverable in 4 PDFs + 1 bonus unlocked after 7 days (reduces refunds), section-by-section page structure, 5 creative angles in image form. (The 5: curiosity, concrete pain, before/after, authority, urgency, one per image, 1:1 format in the source. Rule adopted on 2026-09-04: **4:5** format and 5 angles chosen from 7, see [`12c`](12c-anatomia-low-ticket.md) §12.7-A and [points 35](pontos-em-aberto.md) and [36](pontos-em-aberto.md).) **The right order:** the spreadsheet delivers the offer → MVP in minutes → test with a small budget → if it doesn't sell, AI analyzes the numbers → if it sells, **wait out the 3 validation days and only then raise the budget by the ROAS ruler in [`03`](03-campanha-bidcap.md) §3.2**: ROAS ≥ 2.5 → +50% · 2.0 to 2.5 → +30% · < 2.0 → +20%. Before validation, don't scale. Never double it ("doubling resets the learning"). *(Rule adopted on 2026-09-24, [points 54 and 56](pontos-em-aberto.md); the source said "30% at a time".)*
 
 ### 12.4 Three safety rules
 
@@ -75,10 +75,10 @@ Self-contained interactive mind map (HTML with the data embedded in a `BLOCKS` a
 | **Encontrar** (Find) | potential market ("if there's no ad running, there's no market"), an offer to model, a specialist | 12.1, 12.2, 9.2 |
 | **Entender** (Understand) | market, audience (pain/desire/awareness, "steal their words" from comments and reviews), a single unified list of the offer's features, a single persona with a name and a routine | 8, 12.5 |
 | **Estruturar** (Structure) | product, irresistible offer, promise, unique mechanisms, deliverables ("the word PDF never appears"), a bonus that knocks down an objection, order bump (3 to 5 active), upsell/downsell | 5, 11.1, 11.4 |
-| **Executar** (Execute) | copy in the order **page, then ad, then VSL**; 9x16 creative with a hook in the first 5 seconds; page with two plans and a cross-sell pop-up; checkout with email only and Pix >75%; Instagram BBF; VSL 90s-2min; ads manager set up before spending | 6.1, 6.2, 11.2, 11.3, 11.4 |
-| **Escalar** (Scale) | separate BM and account; CBO with bid cap at full ticket ("ABO is going away"); 17h with no spend = the offer didn't pass; 3x the ticket with ROI<1 kills it; 3 green days raises up to 50% | 3.7, 11.6 |
+| **Executar** (Execute) | copy in the order **page, then ad, then VSL**; 9x16 creative with a hook in the first 3 seconds; page with two plans and a cross-sell pop-up; checkout with email only and Pix >75%; Instagram BBF; VSL 90s-2min; ads manager set up before spending | 6.1, 6.2, 11.2, 11.3, 11.4 |
+| **Escalar** (Scale) | separate BM and account; CBO with an initial bid cap = break-even × 1.5 ("ABO is going away"); 17h with no spend = the offer didn't pass; a creative is only cut with all 4 conditions of 3.7 together (spend ≥ 3x the Bid Cap); 3 green days = offer validated, and from there the budget rises by the ROAS ruler of 3.2 | 3.2, 3.7, 11.6 |
 
-**Note:** the 50% step after 3 green days diverges from the ROAS-based ruler in 03 §3.2 (the 50% only applies at ROAS ≥ 2.5, point 54); see [open point 56](pontos-em-aberto.md). The bid cap at full ticket and the 3x-ticket cut also have diverging rulers elsewhere: [points 58 and 59](pontos-em-aberto.md).
+**Note:** the source said "bid cap at full ticket", "3x the ticket with ROI<1 kills it" and "3 green days raises up to 50%". Since 2026-09-24 the rulers in [`03`](03-campanha-bidcap.md) apply ([points 56, 58 and 59](pontos-em-aberto.md)).
 
 **The two genuinely new things:**
 

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/12a-mineracao-veredito.md

@@ -89,9 +89,9 @@ Brazilian market names as of 2026-09: consolidated marketplace = Hotmart (9.9% +
 
 Rules that follow: fixed fee is the dominant criterion below R$30 · between R$47-147 it balances against available features (bump, one-click upsell) · above R$97 the percentage matters less, prioritize anti-fraud and approval · **without post-checkout one-click purchase, switch platforms BEFORE building the funnel.**
 
-**Core principle:** never clone, always rebuild. Your own name, identity, copy, deliverable, domain, and monetization. It's the same rule 3 from 12.4. **Exception adopted on 2026-09-24:** the functional structure of a quiz funnel can be cloned, with copy, images and brand swapped out ([`12b`](12b-fabrica-de-ofertas.md) §12.11).
+**Core principle:** never clone, always rebuild. Your own name, identity, copy, deliverable, domain, and monetization. It's the same rule 3 from 12.4. It holds **with no exception**: the quiz funnel mirror in [`12b`](12b-fabrica-de-ofertas.md) §12.11 is study material for rebuilding, not an exception to this rule ([point 65](pontos-em-aberto.md)).
 
-**Note on cloning:** besides being a copy, downloading and cloning the page hands you frozen code you can't edit, and it's clear piracy. Same "never clone" rule as above. The exception is the functional quiz mirror in §12.11 of [`12b`](12b-fabrica-de-ofertas.md), which swaps out all the content before publishing.
+**Note on cloning:** besides being a copy, downloading and cloning the page hands you frozen code you can't edit, and it's clear piracy. Same "never clone" rule as above. The quiz mirror in §12.11 of [`12b`](12b-fabrica-de-ofertas.md) doesn't contradict this: it's for studying structure, flow and question logic, and what goes live is your own funnel built from scratch.
 
 > The source also teaches buying the competitor's offer and requesting a refund to extract the deliverable, treating the cost as research expense. **This project does not do that** — it's kept on record as course material, not as a process step. See [Open point 34](pontos-em-aberto.md).
 
@@ -118,7 +118,7 @@ Sale tickets sit between **R$10 and R$19.90**, exactly the entry range that the
 **What can be reused from this:**
 
 1. **Confirms the 12.7 price ruler** with a real market offer, not a hypothesis.
-2. **Shows anchoring in use**: the stated discount reaches 97% (R$397 → R$10). That's aggressive even for low ticket, and it's worth keeping as a reference for how far the market goes, **not as a template**: striking through a price that was never actually charged is a blocked tactic ([guardrails 48](guardrails.md)).
+2. **Shows anchoring in use**: the stated discount reaches 97% (R$397 → R$10). That's aggressive even for low ticket, and it's worth keeping as a reference for how far the market goes, **not as a template**: a struck-through price only if it was actually charged (condition in [guardrails 48](guardrails.md)).
 3. **It's direct competition** if you plan to sell a creative pack. The 4.10 swipe file and the 4.11 templates cover the same ground.
 
 > Apply the 12.2 rubric and the 12.5 verdict here before treating it as a model: there's no data on how many active ads, how long it's been running, or revenue. **It's a sample, not a validated offer.**

package/src/modules/lowticket/skills/wizz-lowticket-metodologia/knowledge/12b-fabrica-de-ofertas.md

@@ -3,9 +3,9 @@
 > Shard of this base's methodology. Index: [`INDEX.md`](INDEX.md) · Always loaded alongside: [`guardrails.md`](guardrails.md)
 > Where the generic skill contradicts this shard, **the shard wins**, and that must be stated in the response.
 
-**What's here:** `12.7` the offer factory — the agent that produces the improved version, the generic reverse-engineering variant, and the internal structure of the PDFs · `12.11` cloning the functional structure of a quiz funnel and taking it to another market
+**What's here:** `12.7` the offer factory — the agent that produces the improved version, the generic reverse-engineering variant, and the internal structure of the PDFs · `12.11` mirroring a quiz funnel as reference, building your own funnel and taking it to another market
 **Siblings:** radar and rubric in [`12-mineracao-ofertas.md`](12-mineracao-ofertas.md) · verdict in [`12a-mineracao-veredito.md`](12a-mineracao-veredito.md) · page anatomy in [`12c-anatomia-low-ticket.md`](12c-anatomia-low-ticket.md)
-**Note:** synthetic testimonials is a blocked tactic — see [guardrails 23](guardrails.md).
+**Note:** testimonials only from real customers; see the condition in [guardrails 23](guardrails.md).
 
 ### 12.7 Offer factory — the agent that produces the improved version
 
@@ -15,7 +15,7 @@ A direct continuation of the offer radar, from the same author (the source autho
 
 > The premise: don't clone the competitor's offer, produce a better version of it.
 
-It's not rhetoric: it's built into the mechanism. The first prompt extracts the **gaps** in the original offer (what that person failed to do), and the page prompt attacks exactly that gap. The modeled offer becomes diagnostic raw material, not a copy template. Consistent with rule 3 from 12.4 and with the "never clone, always rebuild" principle from 12.5 (exception decided on 2026-09-24: the functional structure of a quiz funnel, §12.11 below).
+It's not rhetoric: it's built into the mechanism. The first prompt extracts the **gaps** in the original offer (what that person failed to do), and the page prompt attacks exactly that gap. The modeled offer becomes diagnostic raw material, not a copy template. Consistent with rule 3 from 12.4 and with the "never clone, always rebuild" principle from 12.5 (no exception: the quiz mirror in §12.11 below is reference for rebuilding, [point 65](pontos-em-aberto.md)).
 
 **Setup:**
 
@@ -50,9 +50,9 @@ It's not rhetoric: it's built into the mechanism. The first prompt extracts the
 - **Anti-block footer elements**: legal and institutional text at the bottom of the page to reduce the risk of a Facebook block.
 - **Ban the em dash from the copy.** A finding from human review, not from the prompt: an em dash "gives away that it was made with AI." Adds to the banned-vocabulary list from 7.4.
 - **Free hosting on Vercel**, instead of HostGator/Registro.br. Netlify (drag-and-drop HTML) fills the same role: equivalents for a static LP.
-- **Video creative: do it by hand at first.** For those with little cash, video cuts + free ElevenLabs narration beat setting up the paid stack (Higgsfield, HeyGen, Dream Face) — the combined cost isn't justified before the offer is validated. A useful counterweight to sections 4.6 and 4.9. **Caveat:** the source talks about "grabbing a cut from TikTok"; downloading someone else's TikTok video is a blocked tactic ([guardrails 45](guardrails.md)). The cuts come from your own material, AI-generated footage, or a licensed stock library.
+- **Video creative: do it by hand at first.** For those with little cash, video cuts + free ElevenLabs narration beat setting up the paid stack (Higgsfield, HeyGen, Dream Face) — the combined cost isn't justified before the offer is validated. A useful counterweight to sections 4.6 and 4.9. **Caveat:** the source talks about "grabbing a cut from TikTok"; someone else's TikTok video goes in only as a reference for recording your own (condition in [guardrails 45](guardrails.md)). The cuts that get uploaded come from your own material, AI-generated footage, or a licensed stock library.
 
-> ⚠️ **Synthetic testimonial — recorded, not adopted.** The source suggests filling in social proof with an AI-generated face plus a fictional name and city. This is a **fabricated testimonial**: besides violating Meta's ad policy, it's deceptive advertising (CDC in Brazil, and even stricter in the EU, where 4.8 already requires framing testimonials as individual results). It's kept on record because it's in the material, but **it is not adopted as practice**. See [guardrails 23](guardrails.md).
+> ⚠️ **Testimonials: real customers only.** The source suggests filling in social proof with an AI-generated face plus a fictional name and city. Testimonial structure and scripting still hold; inventing the customer does not, because a fabricated testimonial violates Meta's ad policy and is deceptive advertising (CDC in Brazil, and even stricter in the EU, where 4.8 requires framing testimonials as individual results). Condition in [guardrails 23](guardrails.md).
 
 **The prompts.** They are never read aloud in the class: they only appear written on screen, and were captured via screenshot on 2026-08-22. The file has the literal text of what was visible.
 
@@ -124,15 +124,15 @@ Fixed structure per part: **the part's promise → numbered step-by-step → fin
 
 **Project file convention:** one file per stage (intelligence sheet, positioning, deliverables, page), zero-padded numbering, folder synced with the project's drive.
 
-### 12.11 Cloning a quiz funnel and taking it to another market
+### 12.11 Mirroring a quiz funnel as reference and taking it to another market
 
-> Adopted on 2026-09-24 ([open point 55](pontos-em-aberto.md)): cleared for use. What gets cloned is the **functional structure**; the original's content comes out before publishing.
+> Adopted on 2026-09-24 ([points 55 and 65](pontos-em-aberto.md)): "cloning" here never means copying and publishing another operation's funnel. **The downloaded mirror is study material:** it's for reading the structure, the flow and the logic of the questions. **The deliverable is your own funnel** built from that reference, with your own code, copy, images and brand. Nothing from the original goes live. The "never clone, always rebuild" rule of [`12a`](12a-mineracao-veredito.md) holds here with no exception.
 
-**When to use a quiz.** Don't build a quiz from scratch: a quiz that sells is built by a niche specialist (e.g. relationships). For someone starting out, a **sales page** is the lower-error, more reliably scalable path. A quiz comes into play once there's already a validated quiz funnel in another market to model.
+**When to use a quiz.** Don't build a quiz from scratch: a quiz that sells is built by a niche specialist (e.g. relationships). For someone starting out, a **sales page** is the lower-error, more reliably scalable path. A quiz comes into play once there's already a validated quiz funnel in another market to use as reference.
 
 **Market sequence:** validate in Brazil → translate and test in Portugal/Europe → only then Spain or the UK, once local metrics are validated. An offer already validated in one country: take the creatives that work there and translate the page and funnel. A new offer: launch directly on CBO Highest Volume (3.7). Testing in a new market with a low ticket (**R$15-20**) validates the creative while spending little. *(Number and campaign type from an auto-generated transcript, unverified; "CBO Highest Volume" is the adopted reading.)*
 
-**Why clone through automation.** Quiz platforms (Inlead, xQuiz) don't let you export the funnel. The approach uses Claude Code (or Codex) in VS Code with Playwright, guided by a single prompt that does the technical work. **Testing the cloned funnel locally before publishing** is mandatory: a JS error that slips through breaks the whole live funnel.
+**Why mirror through automation.** Quiz platforms (Inlead, xQuiz) don't let you see the funnel from the inside. The approach uses Claude Code (or Codex) in VS Code with Playwright, guided by a single prompt that does the technical work of building a local mirror for study. **Testing your own funnel locally before publishing** is mandatory: a JS error that slips through breaks the whole live funnel.
 
 **The prompt, in 6 phases (0 to 5; phase 5 is the paragraph right below the table):**
 
@@ -146,6 +146,6 @@ Fixed structure per part: **the part's promise → numbered step-by-step → fin
 
 **Phase 5 · What the mirror doesn't bring over** (list in order of what blocks the funnel): the backend (where the funnel dies without it, usually on quiz submission), a fixed base path hardcoded across several files (if the import resolver isn't fixed, the page goes blank), the payment gateway and product IDs, remote configuration (flags fall back to default and upsells disappear), pixels with real IDs in the code, legal text.
 
-**Before publishing, swap out everything that belongs to the original**: copy, images, logos, brand, legal text and the support email. Also remove the **original owner's pixels, session replay and error tracking**, which would otherwise keep sending your leads' data to their account, along with any hardcoded payment IDs. What's worth keeping is the flow: question order, branches, where the offer appears, the price.
+**What to build from scratch before publishing:** the funnel's code, copy, images, logos, brand, legal text and support email, plus your own pixels and payment IDs. The mirror is never published or edited to go live: if it were, it would carry the **original owner's pixels, session replay and error tracking**, which send the leads' data to their account, and the hardcoded payment IDs. What you take from the mirror is only the reference: question order, branches, where the offer appears, the price.
 
-**Deliverable:** a ready-to-serve site folder plus the mirroring, serving, crawling and asset-audit scripts, per-branch captures, and a README describing what breaks if nothing changes. **Self-check:** zero JS errors, zero missing files, number of steps walked, every branch verified, deep routes returning 200, and nothing about backend, payment or pixels claimed without being backed by a file.
+**Deliverable (the study mirror, not the funnel you publish):** a site folder ready to serve locally plus the mirroring, serving, crawling and asset-audit scripts, per-branch captures, and a README describing what breaks if nothing changes. **Self-check:** zero JS errors, zero missing files, number of steps walked, every branch verified, deep routes returning 200, and nothing about backend, payment or pixels claimed without being backed by a file.