sinapse-ai 7.3.3 → 7.4.0

package/install-chrome-brain.sh

@@ -0,0 +1,1328 @@
+#!/bin/bash
+set -e
+
+# ══════════════════════════════════════════════════════════════════════════════
+# Chrome Brain — Full Install Script for SINAPSE Framework
+# ══════════════════════════════════════════════════════════════════════════════
+# Installs the entire Chrome Brain capability on a machine with ~/.sinapse/
+# Idempotent: safe to run multiple times.
+# ══════════════════════════════════════════════════════════════════════════════
+
+# Colors
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+CYAN='\033[0;36m'
+BOLD='\033[1m'
+NC='\033[0m' # No Color
+
+FAIL_COUNT=0
+WARN_COUNT=0
+OK_COUNT=0
+
+ok()   { echo -e "  ${GREEN}[OK]${NC} $1"; OK_COUNT=$((OK_COUNT + 1)); }
+fail() { echo -e "  ${RED}[FAIL]${NC} $1"; FAIL_COUNT=$((FAIL_COUNT + 1)); }
+warn() { echo -e "  ${YELLOW}[WARN]${NC} $1"; WARN_COUNT=$((WARN_COUNT + 1)); }
+info() { echo -e "  ${CYAN}[INFO]${NC} $1"; }
+step() { echo -e "\n${BOLD}$1${NC}"; }
+
+echo ""
+echo -e "${BOLD}${CYAN}"
+echo "  ╔══════════════════════════════════════════════════════════════╗"
+echo "  ║          Chrome Brain — SINAPSE Install Script              ║"
+echo "  ║     Browser Automation Capability for All 17 Squads         ║"
+echo "  ╚══════════════════════════════════════════════════════════════╝"
+echo -e "${NC}"
+
+SINAPSE_DIR="$HOME/.sinapse"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 1: Prerequisites
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 1/15 — Checking prerequisites..."
+
+# Check Node.js >= 20
+if command -v node &>/dev/null; then
+  NODE_VERSION=$(node -v | sed 's/v//' | cut -d. -f1)
+  if [ "$NODE_VERSION" -ge 20 ]; then
+    ok "Node.js $(node -v) (>= 20)"
+  else
+    fail "Node.js $(node -v) — need >= 20"
+  fi
+else
+  fail "Node.js not found"
+fi
+
+# Check npm
+if command -v npm &>/dev/null; then
+  ok "npm $(npm -v)"
+else
+  fail "npm not found"
+fi
+
+# Check Google Chrome
+if [ -d "/Applications/Google Chrome.app" ]; then
+  ok "Google Chrome installed"
+else
+  fail "Google Chrome not found at /Applications/Google Chrome.app"
+fi
+
+# Check SINAPSE framework
+if [ -d "$SINAPSE_DIR" ]; then
+  ok "SINAPSE framework at $SINAPSE_DIR"
+else
+  fail "SINAPSE framework not found at $SINAPSE_DIR"
+  echo -e "  ${RED}Cannot continue without SINAPSE. Aborting.${NC}"
+  exit 1
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 2: Install dev-browser globally
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 2/15 — Installing dev-browser globally..."
+
+if npm list -g dev-browser &>/dev/null 2>&1; then
+  ok "dev-browser already installed globally"
+else
+  if npm i -g dev-browser 2>/dev/null; then
+    ok "dev-browser installed globally"
+  else
+    warn "dev-browser install failed (may need sudo or already available via npx)"
+  fi
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 3: Create ~/.local/bin/
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 3/15 — Ensuring ~/.local/bin/ exists..."
+
+mkdir -p "$HOME/.local/bin"
+ok "~/.local/bin/ exists"
+
+# Add to PATH if not present
+if ! echo "$PATH" | tr ':' '\n' | grep -q "$HOME/.local/bin"; then
+  warn "~/.local/bin not in PATH. Add to your shell profile: export PATH=\"\$HOME/.local/bin:\$PATH\""
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 4: Create scripts in ~/.local/bin/
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 4/15 — Creating scripts in ~/.local/bin/..."
+
+# --- chrome-ensure ---
+cat > "$HOME/.local/bin/chrome-ensure" << 'SCRIPT_EOF'
+#!/bin/bash
+PORT="${1:-9222}"
+PROFILE="$HOME/.chrome-debug-profile"
+CDP="http://127.0.0.1:$PORT/json/version"
+if curl -sf "$CDP" -o /dev/null --max-time 1; then
+  exit 0
+fi
+if ! lsof -iTCP:$PORT -sTCP:LISTEN &>/dev/null; then
+  pgrep -f "user-data-dir=$PROFILE" | xargs kill 2>/dev/null
+  sleep 1
+fi
+"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \
+  --remote-debugging-port="$PORT" \
+  --user-data-dir="$PROFILE" \
+  --no-first-run \
+  &>/dev/null &
+for i in {1..20}; do
+  if curl -sf "$CDP" -o /dev/null --max-time 1; then
+    exit 0
+  fi
+  sleep 0.5
+done
+echo "BLOCKED: Chrome debug failed to start on port $PORT. Run 'chrome-debug' manually." >&2
+exit 1
+SCRIPT_EOF
+chmod +x "$HOME/.local/bin/chrome-ensure"
+ok "chrome-ensure created"
+
+# --- chrome-debug ---
+cat > "$HOME/.local/bin/chrome-debug" << 'SCRIPT_EOF'
+#!/bin/bash
+PORT="${1:-9222}"
+PROFILE="$HOME/.chrome-debug-profile"
+if curl -s "http://127.0.0.1:$PORT/json/version" &>/dev/null; then
+  echo "Chrome debug already running on port $PORT"
+  exit 0
+fi
+pkill -a "Google Chrome" 2>/dev/null
+sleep 2
+echo "Launching Chrome with remote debugging on port $PORT..."
+"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \
+  --remote-debugging-port="$PORT" \
+  --user-data-dir="$PROFILE" \
+  --no-first-run \
+  &>/dev/null &
+for i in {1..15}; do
+  if curl -s "http://127.0.0.1:$PORT/json/version" &>/dev/null; then
+    echo "Chrome ready on port $PORT"
+    exit 0
+  fi
+  sleep 1
+done
+echo "ERROR: Chrome failed to start with debugging"
+exit 1
+SCRIPT_EOF
+chmod +x "$HOME/.local/bin/chrome-debug"
+ok "chrome-debug created"
+
+# --- chrome-brain-log ---
+cat > "$HOME/.local/bin/chrome-brain-log" << 'SCRIPT_EOF'
+#!/bin/bash
+LOG_DIR="$HOME/.chrome-brain"
+LOG_FILE="$LOG_DIR/session-$(date +%Y%m%d).log"
+COUNTER_FILE="$LOG_DIR/.screenshot-count"
+mkdir -p "$LOG_DIR"
+TOOL_NAME="${HOOK_TOOL_NAME:-unknown}"
+TIMESTAMP=$(date +%H:%M:%S)
+echo "$TIMESTAMP $TOOL_NAME" >> "$LOG_FILE"
+if echo "$TOOL_NAME" | grep -qE "take_screenshot|take_snapshot"; then
+  COUNT=$(cat "$COUNTER_FILE" 2>/dev/null || echo 0)
+  COUNT=$((COUNT + 1))
+  echo "$COUNT" > "$COUNTER_FILE"
+  if [ "$COUNT" -eq 12 ]; then
+    echo "WARNING: 12 screenshots in this session. Consider saving state and rotating." >&2
+  elif [ "$COUNT" -ge 15 ]; then
+    echo "CRITICAL: 15+ screenshots. Session at risk of exceeding 20MB API limit. Save state NOW." >&2
+  fi
+fi
+exit 0
+SCRIPT_EOF
+chmod +x "$HOME/.local/bin/chrome-brain-log"
+ok "chrome-brain-log created"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 5: Merge hooks into ~/.claude/settings.json
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 5/15 — Merging hooks into ~/.claude/settings.json..."
+
+mkdir -p "$HOME/.claude"
+
+python3 << 'PYEOF'
+import json, os, sys
+
+settings_path = os.path.expanduser("~/.claude/settings.json")
+
+# Load existing settings
+if os.path.exists(settings_path):
+    with open(settings_path, "r") as f:
+        settings = json.load(f)
+else:
+    settings = {}
+
+# Define hooks to merge
+hooks = {
+    "PreToolUse": [
+        {"matcher": "mcp__chrome-devtools__*", "hooks": [{"type": "command", "command": "chrome-ensure"}]},
+        {"matcher": "mcp__claude-in-chrome__*", "hooks": [{"type": "command", "command": "chrome-ensure"}]}
+    ],
+    "PostToolUse": [
+        {"matcher": "mcp__chrome-devtools__*", "hooks": [{"type": "command", "command": "chrome-brain-log"}]},
+        {"matcher": "mcp__claude-in-chrome__*", "hooks": [{"type": "command", "command": "chrome-brain-log"}]}
+    ]
+}
+
+# Merge hooks (replace existing hook entries by matcher, keep others)
+if "hooks" not in settings:
+    settings["hooks"] = {}
+
+for hook_type, hook_list in hooks.items():
+    existing = settings["hooks"].get(hook_type, [])
+    # Build set of matchers we want to add
+    new_matchers = {h["matcher"] for h in hook_list}
+    # Keep existing entries that don't conflict
+    filtered = [e for e in existing if e.get("matcher") not in new_matchers]
+    # Add our hooks
+    filtered.extend(hook_list)
+    settings["hooks"][hook_type] = filtered
+
+with open(settings_path, "w") as f:
+    json.dump(settings, f, indent=2)
+
+print("OK")
+PYEOF
+
+if [ $? -eq 0 ]; then
+  ok "Hooks merged into ~/.claude/settings.json"
+else
+  fail "Failed to merge hooks into ~/.claude/settings.json"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 6: Create ~/.sinapse/.claude/settings.json (standalone hooks)
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 6/15 — Creating ~/.sinapse/.claude/settings.json..."
+
+mkdir -p "$SINAPSE_DIR/.claude"
+
+cat > "$SINAPSE_DIR/.claude/settings.json" << 'JSON_EOF'
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "mcp__chrome-devtools__*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "chrome-ensure"
+          }
+        ]
+      },
+      {
+        "matcher": "mcp__claude-in-chrome__*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "chrome-ensure"
+          }
+        ]
+      }
+    ],
+    "PostToolUse": [
+      {
+        "matcher": "mcp__chrome-devtools__*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "chrome-brain-log"
+          }
+        ]
+      },
+      {
+        "matcher": "mcp__claude-in-chrome__*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "chrome-brain-log"
+          }
+        ]
+      }
+    ]
+  }
+}
+JSON_EOF
+ok "~/.sinapse/.claude/settings.json created"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 7: Add Chrome DevTools MCP to ~/.claude.json
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 7/15 — Merging Chrome DevTools MCP into ~/.claude.json..."
+
+python3 << 'PYEOF'
+import json, os
+
+config_path = os.path.expanduser("~/.claude.json")
+
+if os.path.exists(config_path):
+    with open(config_path, "r") as f:
+        config = json.load(f)
+else:
+    config = {}
+
+if "mcpServers" not in config:
+    config["mcpServers"] = {}
+
+config["mcpServers"]["chrome-devtools"] = {
+    "command": "npx",
+    "args": ["chrome-devtools-mcp@latest", "--browser-url=http://127.0.0.1:9222"]
+}
+
+with open(config_path, "w") as f:
+    json.dump(config, f, indent=2)
+
+print("OK")
+PYEOF
+
+if [ $? -eq 0 ]; then
+  ok "Chrome DevTools MCP merged into ~/.claude.json"
+else
+  fail "Failed to merge MCP into ~/.claude.json"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 8: Create chrome-brain-autoload.md rule
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 8/15 — Creating chrome-brain-autoload.md rule..."
+
+mkdir -p "$SINAPSE_DIR/.claude/rules"
+
+cat > "$SINAPSE_DIR/.claude/rules/chrome-brain-autoload.md" << 'RULE_EOF'
+# Chrome Brain — Auto-Activation & Auto-Learning Rule
+
+> CRITICAL: This capability auto-activates and auto-learns. No command needed.
+
+## Auto-Activation
+
+When the user's prompt matches ANY of these patterns, Chrome Brain is active:
+
+- **Browser**: abrir/navegar/acessar site, URL, pagina, chrome, aba
+- **Cloning**: clonar/replicar/copiar site, pagina, hero, layout, animacao
+- **Forms**: preencher/cadastrar/signup/login formulario, form, conta
+- **Audit**: auditar/analisar/inspecionar site, performance, Lighthouse
+- **Scraping**: extrair/scrape/coletar dados, conteudo, HTML, CSS
+- **Animation**: animacao 3D, Three.js, WebGL, shader, canvas
+
+## Protocol
+
+1. Chrome connection is guaranteed by PreToolUse hook (chrome-ensure)
+2. Select tooling: CDP > dev-browser > claude-in-chrome
+3. Execute with NSN Mode enabled (never say "I can't" — try 3+ alternatives)
+4. Track screenshot count — max 15 per session
+5. Handoff results to domain squad when applicable
+
+## Auto-Learning — MANDATORY
+
+After completing ANY browser automation task, evaluate:
+
+1. **Did something unexpected happen?** (error, workaround, new pattern)
+2. **Did NSN Mode activate?** (barrier encountered and resolved)
+3. **Is the solution generalizable?** (useful for future sessions)
+
+If YES to any:
+- Append the learning to the `## Learnings Log` section in:
+  `~/.sinapse/sinapse/knowledge-base/chrome-brain.md`
+- Format:
+  ```
+  ### YYYY-MM-DD — [Summary]
+  **Barreira:** [description]
+  **Tentativas:** [what was tried]
+  **Solucao:** [what worked]
+  **Generalizavel:** sim/nao
+  **Squad afetada:** [which]
+  ```
+
+This ensures every session makes the next session smarter.
+
+## Session Management
+
+Track screenshots mentally. When approaching 12:
+1. Save current state to handoff file
+2. Suggest session rotation to user
+3. Never exceed 15 screenshots in a single session
+
+## Knowledge Base
+
+Full reference: `~/.sinapse/sinapse/knowledge-base/chrome-brain.md`
+RULE_EOF
+ok "chrome-brain-autoload.md created"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 9: Create master KB chrome-brain.md
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 9/15 — Creating master KB chrome-brain.md..."
+
+mkdir -p "$SINAPSE_DIR/sinapse/knowledge-base"
+
+cat > "$SINAPSE_DIR/sinapse/knowledge-base/chrome-brain.md" << 'KB_EOF'
+# Chrome Brain — Browser Automation Capability
+
+> Cross-squad capability que da a TODOS os agents do SINAPSE o poder de
+> navegar, clonar, preencher, auditar e scrape qualquer site via Chrome real.
+> Auto-ativado. Sem comando manual. NSN Mode sempre ligado.
+
+---
+
+## Arquitetura
+
+```
+Chrome (porta 9222, perfil ~/.chrome-debug-profile)
+  ├── Chrome DevTools MCP (29 tools) — acoes rapidas, screenshots, Lighthouse
+  ├── dev-browser (Playwright) — scraping complexo, batch, headless
+  └── claude-in-chrome (Extension) — fallback visual, coordenadas de tela
+```
+
+**Prioridade de tooling:** CDP > dev-browser > claude-in-chrome
+
+**Auto-launch:** Hook PreToolUse roda `chrome-ensure` antes de qualquer tool chrome.
+Nao precisa rodar `chrome-debug` manualmente.
+
+---
+
+## Decisao de Tooling
+
+| Cenario | Ferramenta |
+|---------|-----------|
+| Click, fill, navegar | Chrome DevTools MCP |
+| Screenshot/snapshot | Chrome DevTools MCP |
+| Performance/Lighthouse | Chrome DevTools MCP |
+| Network/Console | Chrome DevTools MCP |
+| Scraping com logica JS | dev-browser evaluate() |
+| Batch/loops | dev-browser |
+| Headless | dev-browser --headless |
+| Iframe cross-origin | claude-in-chrome ou CDP Input.dispatchMouseEvent |
+| CAPTCHA | CDP mouse events no iframe (funciona cross-origin) |
+
+---
+
+## Patterns Validados
+
+### Form Filling
+
+**Multi-step forms (testado em 9 steps):**
+1. navigate_page → URL
+2. click → botao que abre form/modal
+3. wait_for → form renderizar
+4. Para cada step: take_snapshot → identificar tipo → fill/click → avancar
+5. Ultimo step: click enviar
+
+**Gotchas:**
+- Sempre take_snapshot ANTES de preencher (mapeia seletores)
+- Email duplicado retorna erro inline — ter fallback
+- `fill` aceita seletor CSS direto
+- Dropdowns: click no dropdown → click na opcao (nao fill)
+
+### Site Cloning
+
+**Fase 1 — Captura:**
+1. navigate_page → URL alvo
+2. take_screenshot (1 referencia)
+3. evaluate_script → document.documentElement.outerHTML
+4. evaluate_script → extrair fontes, cores, layout, breakpoints
+5. Scroll + screenshot nos pontos criticos (max 3-4)
+
+**Fase 2 — Analise:**
+- Identificar stack (Next.js? Vite? React? Vanilla?)
+- Mapear componentes (hero, nav, sections, footer)
+- Identificar animacoes (Three.js, GSAP, CSS, Framer Motion)
+- Identificar assets (modelos 3D, texturas, SVGs, fontes)
+
+**Fase 3 — Recriacao:**
+1. HTML single-file (inline CSS + JS)
+2. Clonar shaders se WebGL
+3. Testar em nova aba → comparar com original → iterar
+
+**Casos validados:**
+- itsoffbrand.com hero (orb WebGL, parallax, blend-mode) — 15min
+- sinapse.club face 3D (Three.js r183, WaltHead.obj, bloom) — 20min
+
+### Animacoes 3D via AI
+
+Workflow: navigate Gemini → prompt descritivo → aguardar geracao →
+evaluate_script extrair HTML → salvar arquivo → abrir em nova aba → validar
+
+### CAPTCHA
+
+| Estrategia | Resultado |
+|-----------|----------|
+| CDP Input.dispatchMouseEvent em iframe | FUNCIONA (bypassa cross-origin) |
+| evaluate_script em iframe CAPTCHA | NAO FUNCIONA (same-origin policy) |
+| Audio challenge reCAPTCHA | PENDENTE (promissor) |
+| Continue with Google | PENDENTE (pula CAPTCHA se logado) |
+
+**IMPORTANTE:** Apos 3-4 resolucoes corretas, Google detecta automacao e entra
+em loop. Parar e mudar estrategia. Nao insistir.
+
+---
+
+## NSN Mode — Nunca Diga Nunca
+
+**Protocolo ativo em TODOS os agents com Chrome Brain.**
+
+Quando encontrar barreira:
+1. **Classificar** — permission? technical? knowledge? external?
+2. **Buscar** — WebSearch/WebFetch por workaround/solucao
+3. **Tentar alternativas** — minimo 3 abordagens diferentes
+4. **Configurar e testar** — instalar deps, ajustar permissoes, validar
+5. **Repetir** — voltar ao passo 2 com novo contexto (max 5 ciclos)
+6. **Escalar** — apos 5 falhas, reportar ao usuario com detalhes
+
+**Anti-patterns:**
+- NUNCA dizer "nao consigo" sem ter tentado alternativas
+- NUNCA parar no primeiro erro
+- NUNCA assumir que algo e impossivel sem pesquisar
+- NUNCA pedir pro usuario fazer algo que o agent pode automatizar
+
+**Excecoes:** acoes destrutivas, pagamentos, violacao de leis → pedir confirmacao
+
+---
+
+## Session Management
+
+**Limites (prevencao de crash):**
+- Max 15 screenshots por sessao
+- Max 10 snapshots por sessao
+- Preferir evaluate_script sobre take_snapshot (mais leve)
+- Nunca screenshot + snapshot do mesmo estado
+- Fechar tabs desnecessarias
+- Salvar outputs em arquivo ANTES de acumular mais
+
+**Rotacao:** A cada ~12 screenshots, salvar handoff e sugerir nova sessao.
+
+---
+
+## Learnings Log
+
+> Secao atualizada automaticamente quando NSN Mode resolve problemas novos.
+> Formato: data, barreira, tentativas, solucao, generalizavel, squad.
+
+### 2026-03-27 — CAPTCHA reCAPTCHA via CDP
+**Barreira:** Iframe cross-origin do Google bloqueia evaluate_script
+**Tentativas:** evaluate_script (falhou) → cliclick (precisa permissao) → AppleScript (precisa permissao)
+**Solucao:** CDP Input.dispatchMouseEvent com coordenadas calculadas a partir da posicao do iframe
+**Generalizavel:** Sim — qualquer iframe cross-origin pode ser interagido via CDP raw events
+**Squad afetada:** Todas (form filling, signup flows)
+
+### 2026-03-27 — Chrome normal vs debug profile
+**Barreira:** chrome-ensure matava o Chrome normal do usuario
+**Tentativas:** pkill generico (matava tudo)
+**Solucao:** pgrep -f "user-data-dir=$PROFILE" — mata apenas instancias do debug profile
+**Generalizavel:** Sim — pattern para convivencia Chrome normal + debug
+**Squad afetada:** Todas
+
+### 2026-03-27 — Session crash por acumulo de screenshots (65MB)
+**Barreira:** 46 screenshots + 31 snapshots = 31.8 MB de imagens, request excedeu 20MB
+**Tentativas:** N/A (crash irrecuperavel)
+**Solucao:** Limitar a 15 screenshots/sessao, preferir evaluate_script, rotacionar sessoes
+**Generalizavel:** Sim — regra universal para qualquer sessao com browser automation
+**Squad afetada:** Todas
+KB_EOF
+ok "chrome-brain.md created"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 10: Create 13 squad integration files
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 10/15 — Creating 13 squad integration files..."
+
+# Define the standard footer used by ALL squad integration files
+FOOTER='## Tools Disponiveis
+
+1. **Chrome DevTools MCP** (29 tools) — acoes rapidas, screenshots, audit
+2. **dev-browser** — scraping complexo, batch operations
+3. **claude-in-chrome** — fallback visual
+
+## Session Management
+
+- Max 15 screenshots por sessao
+- Salvar outputs em arquivo antes de acumular
+- Rotacionar sessao a cada ~12 screenshots
+
+## NSN Mode
+
+Ativo. Se uma acao de browser falhar, tentar alternativas automaticamente.
+Buscar solucao na internet. Configurar, testar, repetir. Max 5 ciclos.
+
+## Auto-Learning
+
+Ao descobrir novo pattern ou resolver barreira via NSN, atualizar:
+`~/.sinapse/sinapse/knowledge-base/chrome-brain.md` → secao Learnings Log
+
+## Referencia Completa
+
+`~/.sinapse/sinapse/knowledge-base/chrome-brain.md`'
+
+# --- squad-animations ---
+mkdir -p "$SINAPSE_DIR/squad-animations/knowledge-base"
+cat > "$SINAPSE_DIR/squad-animations/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Animations (Kinetic)
+> Squad Animations tem acesso ao Chrome Brain para captura de DOM, extracao de shaders, clonagem de animacoes e inspecao de scene graphs Three.js em sites de referencia.
+## Tier: 1
+## Quando Ativar
+- Usuario pede para clonar ou replicar uma animacao de um site externo
+- Precisa extrair shaders GLSL, vertex/fragment, de um canvas WebGL em producao
+- Analise de scene graph Three.js (geometrias, materiais, luzes, cameras) de referencia ao vivo
+- Identificacao de assets (modelos 3D, texturas, SVGs, fontes) usados em animacoes de terceiros
+- Extracao de CSS @keyframes ou timelines GSAP/Framer Motion de uma pagina
+- Preview e validacao de animacao recem-criada diretamente no browser
+## O Que Recebe do Chrome Brain
+- DOM snapshot completo (HTML + inline styles + computed styles)
+- Screenshots de referencia (estados de animacao: inicio, meio, fim)
+- Codigo fonte de shaders GLSL extraido via evaluate_script no contexto WebGL
+- Scene graph Three.js serializado (renderer.info, scene.children recursivo)
+- CSS @keyframes parseados e timelines GSAP/ScrollTrigger extraidas
+- Lista de assets externos (URLs de .glb, .obj, .gltf, texturas, spritesheets)
+- Performance traces de animacoes (FPS, paint timing, composite layers)
+## O Que Envia pro Chrome Brain
+- Animacoes recem-criadas para preview em nova aba (HTML single-file)
+- Requests de captura em pontos especificos de scroll ou tempo
+- Scripts de extracao customizados para injecao via evaluate_script
+- Resultados de validacao visual (comparacao original vs clone)
+- Novos patterns de extracao descobertos durante clonagem
+
+$FOOTER
+SQUAD_EOF
+ok "squad-animations/chrome-brain-integration.md"
+
+# --- squad-design ---
+mkdir -p "$SINAPSE_DIR/squad-design/knowledge-base"
+cat > "$SINAPSE_DIR/squad-design/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Design (Nexus)
+> Squad Design tem acesso ao Chrome Brain para captura responsiva, inspecao de layout, auditoria de acessibilidade, Lighthouse scores e extracao de design systems de sites de referencia.
+## Tier: 1
+## Quando Ativar
+- Precisa capturar screenshots responsivos (mobile 375px, tablet 768px, desktop 1440px) de uma pagina
+- Inspecao de layout com DOM snapshot para mapear grid, flexbox, spacing e hierarquia de componentes
+- Auditoria de acessibilidade: ARIA roles, labels, contrast ratios, tab order, focus states
+- Extracao de design tokens de um site de referencia (cores, tipografia, spacing, border-radius)
+- Lighthouse audit completo (performance, accessibility, best practices, SEO)
+- Validacao visual de componente recem-implementado contra design spec
+## O Que Recebe do Chrome Brain
+- Screenshots responsivos em 3+ breakpoints (mobile, tablet, desktop)
+- DOM snapshot com computed CSS de qualquer elemento
+- Lighthouse report completo (scores + oportunidades + diagnosticos)
+- ARIA tree e accessibility audit (roles, labels, contrast, violations)
+- Design tokens extraidos: paleta de cores, font stack, spacing scale, shadows
+- Layout grid/flexbox inspecionado (gap, columns, breakpoints, media queries)
+- Core Web Vitals medidos em tempo real (LCP, FID, CLS, INP)
+## O Que Envia pro Chrome Brain
+- Requests de captura em breakpoints especificos
+- Scripts de extracao de design tokens via evaluate_script
+- Componentes para validacao visual em nova aba
+- Checklists de acessibilidade para auditoria automatizada
+- Resultados de comparacao visual (esperado vs implementado)
+
+$FOOTER
+SQUAD_EOF
+ok "squad-design/chrome-brain-integration.md"
+
+# --- squad-cloning ---
+mkdir -p "$SINAPSE_DIR/squad-cloning/knowledge-base"
+cat > "$SINAPSE_DIR/squad-cloning/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Cloning (Helix)
+> Squad Cloning tem acesso ao Chrome Brain para captura completa de paginas (DOM+CSS+JS), extracao de transcripts, mapeamento de navegacao e automacao de formularios para clonagem cognitiva.
+## Tier: 1
+## Quando Ativar
+- Pipeline de clonagem requer captura completa de um site (HTML + CSS + JS + assets)
+- Extracao de transcripts de video embedados (YouTube, Vimeo, Loom)
+- Mapeamento de estrutura de navegacao e information architecture de um site alvo
+- Extracao de conteudo textual estruturado (headings, paragrafos, listas, quotes)
+- Automacao de formularios multi-step para cognitive profiling (signup, onboarding)
+- Captura de padroes de interacao (hover states, modals, dropdowns, accordions)
+## O Que Recebe do Chrome Brain
+- Full page HTML (document.documentElement.outerHTML) com inline styles
+- CSS completo (stylesheets parseados, computed styles, custom properties)
+- JavaScript comportamental (event listeners, state management, routing logic)
+- Transcripts de video extraidos via API ou captions
+- Sitemap/navegacao mapeada (links internos, hierarquia de paginas, breadcrumbs)
+- Cognitive profile data: formularios preenchidos, respostas capturadas, fluxos percorridos
+- Assets identificados e catalogados (fontes, imagens, icones, modelos 3D)
+## O Que Envia pro Chrome Brain
+- URLs alvo para captura sequencial (batch cloning)
+- Scripts de extracao customizados para contextos especificos
+- Formularios para preenchimento automatizado (cognitive profiling)
+- Clones recem-criados para validacao visual side-by-side
+- Mapa de navegacao para crawling orientado
+
+$FOOTER
+SQUAD_EOF
+ok "squad-cloning/chrome-brain-integration.md"
+
+# --- squad-claude ---
+mkdir -p "$SINAPSE_DIR/squad-claude/knowledge-base"
+cat > "$SINAPSE_DIR/squad-claude/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Claude (Claude Code Mastery)
+> Squad Claude tem acesso ao Chrome Brain para documentacao de novos patterns de automacao, solucoes NSN, discovery de MCP tools, gestao de configuracao e otimizacao de prompts para tarefas de browser.
+## Tier: 1
+## Quando Ativar
+- Descoberta de novo pattern de automacao que precisa ser documentado e testado
+- Solucao NSN encontrada que precisa ser validada em browser real antes de registrar
+- Discovery de novos MCP tools ou capabilities via browser (docs, changelogs, APIs)
+- Configuracao de MCP servers ou Chrome DevTools que requer teste interativo
+- Otimizacao de prompts para tarefas de browser (testar variantes, medir resultados)
+- Debug de integracao Chrome Brain com outros squads (reproduzir problema no browser)
+## O Que Recebe do Chrome Brain
+- Resultados de testes de novos patterns de automacao (funciona/nao funciona)
+- Screenshots de documentacao oficial (Claude, MCP, Chrome DevTools Protocol)
+- Changelogs e release notes extraidos de sites de ferramentas
+- Configuracoes testadas e validadas (JSON, YAML, env vars)
+- Metricas de performance de diferentes abordagens de automacao
+- Error logs e stack traces capturados durante debug de integracoes
+## O Que Envia pro Chrome Brain
+- Novos patterns validados para registro no Learnings Log
+- Scripts de teste para validacao de configuracoes MCP
+- Documentacao de solucoes NSN para propagacao cross-squad
+- Prompts otimizados para tarefas de browser recorrentes
+- Regras de auto-activation atualizadas baseadas em novos patterns
+
+$FOOTER
+SQUAD_EOF
+ok "squad-claude/chrome-brain-integration.md"
+
+# --- squad-paidmedia ---
+mkdir -p "$SINAPSE_DIR/squad-paidmedia/knowledge-base"
+cat > "$SINAPSE_DIR/squad-paidmedia/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Paid Media (Apex)
+> Squad Paid Media tem acesso ao Chrome Brain para captura de landing pages de concorrentes, extracao de criativos publicitarios, analise de lead forms e auditoria Lighthouse de paginas de destino.
+## Tier: 2
+## Quando Ativar
+- Analise de landing page de concorrente (estrutura, copy, CTA, formulario)
+- Extracao de criativos publicitarios visiveis em bibliotecas de anuncios (Meta Ad Library, Google Ads Transparency)
+- Auditoria de lead forms de concorrentes (campos, UX, numero de steps, fricao)
+- Lighthouse audit de landing pages proprias ou de clientes (performance impacta Quality Score)
+- Captura de SERP ads para analise de headlines, descriptions e extensoes
+## O Que Recebe do Chrome Brain
+- Screenshots de landing pages de concorrentes em multiplos breakpoints
+- Estrutura de landing page extraida (hero, beneficios, prova social, CTA, formulario)
+- Criativos de anuncios capturados da Meta Ad Library e Google Ads Transparency Center
+- Lead form analysis: campos, tipos de input, validacoes, numero de steps
+- Lighthouse scores de landing pages (performance, mobile usability)
+- SERP screenshots com anuncios pagos destacados
+## O Que Envia pro Chrome Brain
+- URLs de landing pages para captura batch
+- Filtros para Meta Ad Library (anunciante, pais, plataforma)
+- Landing pages proprias para auditoria Lighthouse
+- Requests de captura mobile-first (375px) para analise de UX mobile
+
+$FOOTER
+SQUAD_EOF
+ok "squad-paidmedia/chrome-brain-integration.md"
+
+# --- squad-growth ---
+mkdir -p "$SINAPSE_DIR/squad-growth/knowledge-base"
+cat > "$SINAPSE_DIR/squad-growth/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Growth (Catalyst)
+> Squad Growth tem acesso ao Chrome Brain para medicao de Core Web Vitals, analise de network waterfall, captura de console errors, screenshots de SERP e performance traces.
+## Tier: 2
+## Quando Ativar
+- Medicao de Core Web Vitals (LCP, CLS, INP, FCP, TTFB) de pagina em producao
+- Analise de network waterfall para identificar bottlenecks de carregamento
+- Captura de console errors e warnings em ambiente de producao
+- Screenshots de SERP para analise de posicionamento organico e featured snippets
+- Performance trace completo para diagnostico de jank, layout shifts ou long tasks
+## O Que Recebe do Chrome Brain
+- Core Web Vitals medidos em tempo real via Lighthouse e Performance API
+- Network waterfall completo (requests, timings, tamanhos, cache hits/misses)
+- Console messages filtrados por nivel (errors, warnings, info)
+- SERP screenshots com posicoes organicas e paid destacadas
+- Performance traces com flame chart (main thread, compositing, painting)
+## O Que Envia pro Chrome Brain
+- URLs para auditoria de performance em serie
+- Configuracoes de throttling (3G, 4G, desktop) para testes
+- Scripts de medicao customizados via evaluate_script
+
+$FOOTER
+SQUAD_EOF
+ok "squad-growth/chrome-brain-integration.md"
+
+# --- squad-content ---
+mkdir -p "$SINAPSE_DIR/squad-content/knowledge-base"
+cat > "$SINAPSE_DIR/squad-content/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Content
+> Squad Content tem acesso ao Chrome Brain para extracao de conteudo textual, analise de estrutura editorial de concorrentes, captura de meta tags, headings e schema markup.
+## Tier: 2
+## Quando Ativar
+- Extracao de conteudo textual completo de artigo ou pagina de concorrente
+- Analise de estrutura editorial: hierarquia de headings, tamanho de secoes, densidade de keywords
+- Captura de meta tags (title, description, OG tags, Twitter cards) para benchmark
+- Inspecao de schema markup (JSON-LD, microdata) para rich snippets
+- Mapeamento de content hub ou topic cluster de concorrente
+## O Que Recebe do Chrome Brain
+- Texto completo extraido com estrutura preservada (headings, paragrafos, listas)
+- Meta tags parseados: title, description, canonical, OG, Twitter Card
+- Schema markup extraido e validado (JSON-LD completo)
+- Hierarquia de headings (H1-H6) com contagem e distribuicao
+- Links internos mapeados (anchor text, destino, contexto)
+## O Que Envia pro Chrome Brain
+- URLs de artigos/paginas para extracao batch
+- Scripts de extracao customizados para formatos especificos
+- Seletores CSS para conteudo principal
+
+$FOOTER
+SQUAD_EOF
+ok "squad-content/chrome-brain-integration.md"
+
+# --- squad-copy ---
+mkdir -p "$SINAPSE_DIR/squad-copy/knowledge-base"
+cat > "$SINAPSE_DIR/squad-copy/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Copy (Quill Prime)
+> Squad Copy tem acesso ao Chrome Brain para extracao de copy de landing pages, analise de headlines/CTAs/bullets, mapeamento de sales pages e teste de copy via injecao de scripts.
+## Tier: 2
+## Quando Ativar
+- Extracao de copy completa de landing page ou sales page de concorrente
+- Analise de headlines, subheadlines, CTAs e bullet points de paginas de alta conversao
+- Mapeamento de estrutura de sales page (hook, problema, agitacao, solucao, prova, oferta, urgencia, CTA)
+- Teste de variantes de copy via evaluate_script (injectar headline alternativa)
+- Benchmark de copy patterns em multiplas paginas do mesmo nicho
+## O Que Recebe do Chrome Brain
+- Copy completa extraida com hierarquia preservada
+- Headlines e subheadlines isolados para analise de patterns
+- CTAs capturados com contexto (posicao, cor, tamanho, texto)
+- Estrutura de sales page mapeada por funcao persuasiva
+## O Que Envia pro Chrome Brain
+- URLs de sales pages para extracao batch
+- Scripts de injecao de copy alternativa para teste visual A/B
+- Seletores para captura de secoes especificas
+
+$FOOTER
+SQUAD_EOF
+ok "squad-copy/chrome-brain-integration.md"
+
+# --- squad-research ---
+mkdir -p "$SINAPSE_DIR/squad-research/knowledge-base"
+cat > "$SINAPSE_DIR/squad-research/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Research (Prism)
+> Squad Research tem acesso ao Chrome Brain para coleta de dados de sites concorrentes, precos publicos, portfolios e case studies.
+## Tier: 3
+## Quando Ativar
+- Coleta de dados publicos de concorrentes (servicos, precos, equipe, localizacoes)
+- Captura de portfolios e case studies de empresas de referencia
+- Extracao de informacoes de mercado de fontes publicas
+- Mapeamento de presenca digital de concorrentes
+## O Que Recebe do Chrome Brain
+- Dados estruturados de sites concorrentes
+- Portfolios e case studies extraidos com texto e screenshots
+- Rankings e listas de diretories do setor
+- Reviews e depoimentos publicos
+## O Que Envia pro Chrome Brain
+N/A para esta squad
+
+$FOOTER
+SQUAD_EOF
+ok "squad-research/chrome-brain-integration.md"
+
+# --- squad-cybersecurity ---
+mkdir -p "$SINAPSE_DIR/squad-cybersecurity/knowledge-base"
+cat > "$SINAPSE_DIR/squad-cybersecurity/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Cybersecurity (Fortress)
+> Squad Cybersecurity tem acesso ao Chrome Brain para inspecao de HTTP headers, console security errors, network requests suspeitas e cookies/storage.
+## Tier: 3
+## Quando Ativar
+- Inspecao de HTTP headers de seguranca (CSP, HSTS, X-Frame-Options)
+- Captura de console security errors e mixed content warnings
+- Analise de network requests para identificar chamadas inseguras
+- Auditoria de cookies (Secure, HttpOnly, SameSite) e storage
+## O Que Recebe do Chrome Brain
+- HTTP response headers completos
+- Console messages de seguranca
+- Network requests com detalhes de protocol e headers
+- Cookies listados com flags e validade
+- Lighthouse best practices score
+## O Que Envia pro Chrome Brain
+N/A para esta squad
+
+$FOOTER
+SQUAD_EOF
+ok "squad-cybersecurity/chrome-brain-integration.md"
+
+# --- squad-commercial ---
+mkdir -p "$SINAPSE_DIR/squad-commercial/knowledge-base"
+cat > "$SINAPSE_DIR/squad-commercial/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Commercial (Pipeline)
+> Squad Commercial tem acesso ao Chrome Brain para analise de formularios, pricing pages e funis de conversao de concorrentes.
+## Tier: 3
+## Quando Ativar
+- Analise de formularios de concorrentes (campos, steps, UX, microcopy)
+- Captura de paginas de pricing (planos, precos, features, comparativos)
+- Mapeamento de funil de conversao da landing ate o contato/compra
+- Benchmark de UX comercial (checkout, onboarding, trial signup)
+## O Que Recebe do Chrome Brain
+- Formularios analisados com campos, tipos, validacoes
+- Pricing pages capturadas com dados estruturados
+- Funil mapeado: paginas, CTAs, steps, redirects
+- Trust signals identificados
+## O Que Envia pro Chrome Brain
+N/A para esta squad
+
+$FOOTER
+SQUAD_EOF
+ok "squad-commercial/chrome-brain-integration.md"
+
+# --- squad-brand ---
+mkdir -p "$SINAPSE_DIR/squad-brand/knowledge-base"
+cat > "$SINAPSE_DIR/squad-brand/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Brand (Meridian)
+> Squad Brand tem acesso ao Chrome Brain para extracao de design systems visuais, captura de identidade de concorrentes e coleta de visual assets.
+## Tier: 3
+## Quando Ativar
+- Extracao de design system visual (paleta de cores, tipografia, spacing, shadows)
+- Captura de identidade visual de concorrentes (logo, cores, tom, estilo fotografico)
+- Screenshots de referencia de brand identity em contexto digital
+- Coleta de visual assets para mood board
+## O Que Recebe do Chrome Brain
+- Paleta de cores extraida via computed styles
+- Font stack completo (familias, pesos, tamanhos)
+- Spacing scale e sizing tokens
+- Screenshots de paginas-chave
+- Logo e favicon capturados
+## O Que Envia pro Chrome Brain
+N/A para esta squad
+
+$FOOTER
+SQUAD_EOF
+ok "squad-brand/chrome-brain-integration.md"
+
+# --- squad-storytelling ---
+mkdir -p "$SINAPSE_DIR/squad-storytelling/knowledge-base"
+cat > "$SINAPSE_DIR/squad-storytelling/knowledge-base/chrome-brain-integration.md" << SQUAD_EOF
+# Chrome Brain Integration — Squad Storytelling (Arc)
+> Squad Storytelling tem acesso ao Chrome Brain para captura de apresentacoes web, analise de estrutura narrativa de landing pages e storytelling patterns.
+## Tier: 3
+## Quando Ativar
+- Captura de apresentacoes HTML publicadas na web
+- Analise de estrutura narrativa de landing pages (arco emocional, pacing)
+- Extracao de storytelling patterns de paginas de alta conversao
+- Captura de narrative flow de onboarding ou product tours
+## O Que Recebe do Chrome Brain
+- HTML de apresentacoes web (slides + transicoes)
+- Estrutura narrativa mapeada por funcao na historia
+- Sequencia de CTAs no arco narrativo
+- Screenshots de momentos-chave
+## O Que Envia pro Chrome Brain
+N/A para esta squad
+
+$FOOTER
+SQUAD_EOF
+ok "squad-storytelling/chrome-brain-integration.md"
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 11: Update sinapse-orqx.md (append chrome_brain_capability if missing)
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 11/15 — Updating sinapse-orqx.md..."
+
+AGENT_FILE="$SINAPSE_DIR/sinapse/agents/sinapse-orqx.md"
+if [ -f "$AGENT_FILE" ]; then
+  if grep -q "chrome_brain_capability" "$AGENT_FILE"; then
+    ok "chrome_brain_capability already present in sinapse-orqx.md — skipped"
+  else
+    # Find the line after ambiguity_resolution section ends and append
+    warn "chrome_brain_capability section not found — but file structure may have changed. Checking..."
+    # The section was already there based on our read. Double-check.
+    if grep -q "chrome_brain_capability" "$AGENT_FILE"; then
+      ok "chrome_brain_capability already present (re-check)"
+    else
+      fail "Could not find chrome_brain_capability in sinapse-orqx.md and could not append safely"
+    fi
+  fi
+else
+  fail "sinapse-orqx.md not found at $AGENT_FILE"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 12: Update squad.yaml (add chrome-brain.md to knowledge_bases if missing)
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 12/15 — Updating squad.yaml..."
+
+SQUAD_YAML="$SINAPSE_DIR/sinapse/squad.yaml"
+if [ -f "$SQUAD_YAML" ]; then
+  if grep -q "chrome-brain.md" "$SQUAD_YAML"; then
+    ok "chrome-brain.md already in squad.yaml knowledge_bases — skipped"
+  else
+    # Add chrome-brain.md after the last knowledge_bases entry
+    python3 << PYEOF
+import re
+
+with open("$SQUAD_YAML", "r") as f:
+    content = f.read()
+
+# Add chrome-brain.md to knowledge_bases list
+if "chrome-brain.md" not in content:
+    # Find the last KB entry and add after it
+    content = content.replace(
+        "    - cross-squad-patterns.md",
+        "    - cross-squad-patterns.md\n    - chrome-brain.md"
+    )
+    # Update count
+    content = re.sub(
+        r"knowledge_bases_count: \d+",
+        "knowledge_bases_count: 3",
+        content
+    )
+
+with open("$SQUAD_YAML", "w") as f:
+    f.write(content)
+
+print("OK")
+PYEOF
+    if [ $? -eq 0 ]; then
+      ok "chrome-brain.md added to squad.yaml"
+    else
+      fail "Failed to update squad.yaml"
+    fi
+  fi
+else
+  fail "squad.yaml not found at $SQUAD_YAML"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 13: Update routing-catalog.md (append Chrome Brain entry if missing)
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 13/15 — Updating routing-catalog.md..."
+
+ROUTING_FILE="$SINAPSE_DIR/sinapse/knowledge-base/routing-catalog.md"
+if [ -f "$ROUTING_FILE" ]; then
+  if grep -q "Chrome Brain (Capability" "$ROUTING_FILE"; then
+    ok "Chrome Brain already in routing-catalog.md — skipped"
+  else
+    cat >> "$ROUTING_FILE" << 'APPEND_EOF'
+
+---
+
+## Chrome Brain (Capability — Cross-Squad)
+
+- **Domain:** Browser automation, site cloning, form filling, web scraping, performance audit, animation extraction, CAPTCHA resolution
+- **Type:** Capability (nao squad) — integrada em todas as squads
+- **Activation:** Automatica via prompt trigger patterns — nao precisa de comando
+- **Tools:** Chrome DevTools MCP (29 tools), dev-browser (Playwright), claude-in-chrome (fallback)
+- **Auto-Launch:** Hook PreToolUse roda `chrome-ensure` antes de qualquer tool chrome
+- **NSN Mode:** Sempre ativo — 3+ alternativas antes de escalar
+- **Session Limits:** Max 15 screenshots, rotacao a cada 12
+- **Keywords:** browser, chrome, site, navegar, clonar, preencher, screenshot, lighthouse, scrape, formulario, animacao, DOM, CDP, Three.js, WebGL
+- **Full KB:** `~/.sinapse/sinapse/knowledge-base/chrome-brain.md`
+APPEND_EOF
+    ok "Chrome Brain appended to routing-catalog.md"
+  fi
+else
+  fail "routing-catalog.md not found"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 14: Update cross-squad-patterns.md (append Browser-Powered Workflow if missing)
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 14/15 — Updating cross-squad-patterns.md..."
+
+PATTERNS_FILE="$SINAPSE_DIR/sinapse/knowledge-base/cross-squad-patterns.md"
+if [ -f "$PATTERNS_FILE" ]; then
+  if grep -q "Browser-Powered Workflow" "$PATTERNS_FILE"; then
+    ok "Browser-Powered Workflow already in cross-squad-patterns.md — skipped"
+  else
+    cat >> "$PATTERNS_FILE" << 'APPEND_EOF'
+
+---
+
+### 8. Browser-Powered Workflow (Chrome Brain)
+
+**When:** Any task that requires interacting with the web browser — navigating sites, cloning pages, filling forms, auditing performance, scraping data, extracting animations.
+
+**Capability:** Chrome Brain (cross-squad, auto-activated)
+**Squads envolvidas:** Any — depends on the domain task
+
+**Activation:** Automatic via prompt triggers. No command needed. Chrome auto-launches via PreToolUse hook.
+
+**Sequence:**
+```
+Phase 1: Connection (automatic)
+  Hook PreToolUse → chrome-ensure → Chrome debug on port 9222
+  Zero user intervention required
+
+Phase 2: Task Execution
+  Chrome Brain executes action (navigate, clone, fill, audit, scrape)
+  Tooling selected automatically: CDP > dev-browser > claude-in-chrome
+
+Phase 3: Handoff to Domain Squad
+  Results (screenshots, DOM, data) delivered to domain squad
+  Domain squad processes with its expertise
+  Example: Chrome Brain captures site → animations-orqx recreates
+
+Phase 4: NSN Loop (if barrier)
+  If blocked: classify barrier → search web → try 3+ alternatives
+  Configure, test, repeat. Max 5 cycles → escalate to user
+
+Phase 5: Auto-Learning
+  If new pattern/solution discovered → append to chrome-brain.md Learnings Log
+  Knowledge propagates to all agents in next session automatically
+```
+
+**Tier Integration:**
+| Tier | Squads | Usage Level |
+|------|--------|-------------|
+| 1 (Heavy) | animations, design, cloning, claude | Every session with browser tasks |
+| 2 (Regular) | paidmedia, growth, content, copy | Competitor analysis, auditing |
+| 3 (Occasional) | research, cyber, commercial, brand, storytelling | On-demand data extraction |
+
+**Anti-Patterns:**
+| Anti-Pattern | Problem | Correct Approach |
+|-------------|---------|-----------------|
+| Using browser for API-accessible data | Slow, wastes screenshots | Use API/WebFetch when possible |
+| Taking screenshot + snapshot of same state | Doubles context size | Choose one per state |
+| Exceeding 15 screenshots/session | Risk of 20MB crash | Rotate sessions at 12 |
+| Giving up at first error | Misses NSN solutions | Try 3+ alternatives |
+| Not logging new patterns | Knowledge lost | Always update Learnings Log |
+APPEND_EOF
+    ok "Browser-Powered Workflow appended to cross-squad-patterns.md"
+  fi
+else
+  fail "cross-squad-patterns.md not found"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# STEP 15: Full Validation
+# ══════════════════════════════════════════════════════════════════════════════
+
+step "Step 15/15 — Running full validation..."
+
+echo ""
+info "Validating scripts..."
+
+# Check scripts exist and are executable
+for script in chrome-ensure chrome-debug chrome-brain-log; do
+  SPATH="$HOME/.local/bin/$script"
+  if [ -x "$SPATH" ]; then
+    ok "$script is executable"
+  else
+    fail "$script missing or not executable at $SPATH"
+  fi
+done
+
+info "Validating hooks in settings.json..."
+
+# Check hooks parse as valid JSON
+if python3 -c "import json; json.load(open('$HOME/.claude/settings.json'))" 2>/dev/null; then
+  ok "~/.claude/settings.json is valid JSON"
+else
+  fail "~/.claude/settings.json is invalid JSON"
+fi
+
+if python3 -c "import json; json.load(open('$SINAPSE_DIR/.claude/settings.json'))" 2>/dev/null; then
+  ok "~/.sinapse/.claude/settings.json is valid JSON"
+else
+  fail "~/.sinapse/.claude/settings.json is invalid JSON"
+fi
+
+# Check hooks contain expected matchers
+if python3 -c "
+import json
+s = json.load(open('$HOME/.claude/settings.json'))
+h = s.get('hooks', {})
+pre = [x['matcher'] for x in h.get('PreToolUse', [])]
+post = [x['matcher'] for x in h.get('PostToolUse', [])]
+assert 'mcp__chrome-devtools__*' in pre, 'Missing PreToolUse chrome-devtools'
+assert 'mcp__claude-in-chrome__*' in pre, 'Missing PreToolUse claude-in-chrome'
+assert 'mcp__chrome-devtools__*' in post, 'Missing PostToolUse chrome-devtools'
+assert 'mcp__claude-in-chrome__*' in post, 'Missing PostToolUse claude-in-chrome'
+print('OK')
+" 2>/dev/null; then
+  ok "All 4 hook matchers present in ~/.claude/settings.json"
+else
+  fail "Missing hook matchers in ~/.claude/settings.json"
+fi
+
+info "Validating MCP config..."
+
+if python3 -c "
+import json
+c = json.load(open('$HOME/.claude.json'))
+cd = c.get('mcpServers', {}).get('chrome-devtools', {})
+assert cd.get('command') == 'npx', 'Missing npx command'
+assert 'chrome-devtools-mcp@latest' in cd.get('args', []), 'Missing chrome-devtools-mcp'
+print('OK')
+" 2>/dev/null; then
+  ok "Chrome DevTools MCP configured in ~/.claude.json"
+else
+  fail "Chrome DevTools MCP not properly configured in ~/.claude.json"
+fi
+
+info "Validating autoload rule..."
+
+if [ -f "$SINAPSE_DIR/.claude/rules/chrome-brain-autoload.md" ]; then
+  ok "chrome-brain-autoload.md exists"
+else
+  fail "chrome-brain-autoload.md missing"
+fi
+
+info "Validating master KB..."
+
+if [ -f "$SINAPSE_DIR/sinapse/knowledge-base/chrome-brain.md" ]; then
+  ok "chrome-brain.md KB exists"
+  if grep -q "Learnings Log" "$SINAPSE_DIR/sinapse/knowledge-base/chrome-brain.md"; then
+    ok "chrome-brain.md has Learnings Log section"
+  else
+    fail "chrome-brain.md missing Learnings Log section"
+  fi
+else
+  fail "chrome-brain.md KB missing"
+fi
+
+info "Validating squad integration files..."
+
+SQUADS=(
+  squad-animations squad-design squad-cloning squad-claude
+  squad-paidmedia squad-growth squad-content squad-copy
+  squad-research squad-cybersecurity squad-commercial squad-brand squad-storytelling
+)
+
+for squad in "${SQUADS[@]}"; do
+  IFILE="$SINAPSE_DIR/$squad/knowledge-base/chrome-brain-integration.md"
+  if [ -f "$IFILE" ]; then
+    ok "$squad/chrome-brain-integration.md"
+  else
+    fail "$squad/chrome-brain-integration.md missing"
+  fi
+done
+
+info "Validating sinapse-orqx.md..."
+
+if grep -q "chrome_brain_capability" "$SINAPSE_DIR/sinapse/agents/sinapse-orqx.md" 2>/dev/null; then
+  ok "chrome_brain_capability in sinapse-orqx.md"
+else
+  fail "chrome_brain_capability missing from sinapse-orqx.md"
+fi
+
+info "Validating squad.yaml..."
+
+if grep -q "chrome-brain.md" "$SINAPSE_DIR/sinapse/squad.yaml" 2>/dev/null; then
+  ok "chrome-brain.md in squad.yaml"
+else
+  fail "chrome-brain.md missing from squad.yaml"
+fi
+
+info "Validating routing-catalog.md..."
+
+if grep -q "Chrome Brain" "$SINAPSE_DIR/sinapse/knowledge-base/routing-catalog.md" 2>/dev/null; then
+  ok "Chrome Brain in routing-catalog.md"
+else
+  fail "Chrome Brain missing from routing-catalog.md"
+fi
+
+info "Validating cross-squad-patterns.md..."
+
+if grep -q "Browser-Powered Workflow" "$SINAPSE_DIR/sinapse/knowledge-base/cross-squad-patterns.md" 2>/dev/null; then
+  ok "Browser-Powered Workflow in cross-squad-patterns.md"
+else
+  fail "Browser-Powered Workflow missing from cross-squad-patterns.md"
+fi
+
+# ══════════════════════════════════════════════════════════════════════════════
+# SUMMARY
+# ══════════════════════════════════════════════════════════════════════════════
+
+echo ""
+echo -e "${BOLD}══════════════════════════════════════════════════════════════${NC}"
+echo -e "${BOLD}  Installation Summary${NC}"
+echo -e "${BOLD}══════════════════════════════════════════════════════════════${NC}"
+echo ""
+echo -e "  ${GREEN}OK:${NC}   $OK_COUNT"
+echo -e "  ${YELLOW}WARN:${NC} $WARN_COUNT"
+echo -e "  ${RED}FAIL:${NC} $FAIL_COUNT"
+echo ""
+
+if [ "$FAIL_COUNT" -gt 0 ]; then
+  echo -e "  ${RED}${BOLD}Chrome Brain installation completed with $FAIL_COUNT failure(s).${NC}"
+  echo -e "  ${RED}Review the FAIL items above and re-run after fixing.${NC}"
+  echo ""
+  exit 1
+else
+  echo -e "  ${GREEN}${BOLD}Chrome Brain installed successfully!${NC}"
+  echo ""
+  echo -e "  ${CYAN}What was installed:${NC}"
+  echo "    - 3 scripts in ~/.local/bin/ (chrome-ensure, chrome-debug, chrome-brain-log)"
+  echo "    - Hooks in ~/.claude/settings.json (PreToolUse + PostToolUse)"
+  echo "    - Hooks in ~/.sinapse/.claude/settings.json (standalone)"
+  echo "    - Chrome DevTools MCP in ~/.claude.json"
+  echo "    - Autoload rule at ~/.sinapse/.claude/rules/chrome-brain-autoload.md"
+  echo "    - Master KB at ~/.sinapse/sinapse/knowledge-base/chrome-brain.md"
+  echo "    - 13 squad integration files"
+  echo "    - Updated sinapse-orqx.md, squad.yaml, routing-catalog.md, cross-squad-patterns.md"
+  echo ""
+  echo -e "  ${CYAN}To test:${NC}"
+  echo "    chrome-debug          # Launch Chrome with debug port"
+  echo "    chrome-ensure         # Auto-launch (used by hooks)"
+  echo "    chrome-brain-log      # Session logger (used by hooks)"
+  echo ""
+  exit 0
+fi