oryon-framework 1.0.0

package/.claude/rules/mcp-usage.md

@@ -0,0 +1,176 @@
+---
+paths: **/*
+---
+
+# MCP Server Usage Rules - Oryon Architecture
+
+## MCP Governance
+
+**IMPORTANT:** All MCP infrastructure management is handled EXCLUSIVELY by the **@ship agent**.
+
+| Operation | Agent | Command |
+|-----------|-------|---------|
+| Search MCP catalog | DevOps | `*search-mcp` |
+| Add MCP server | DevOps | `*add-mcp` |
+| List enabled MCPs | DevOps | `*list-mcps` |
+| Remove MCP server | DevOps | `*remove-mcp` |
+| Setup Docker MCP | DevOps | `*setup-mcp-docker` |
+
+Other agents are MCP **consumers**, not administrators. If MCP management is needed, delegate to @ship.
+
+---
+
+## MCP Configuration Architecture
+
+Oryon uses Docker MCP Toolkit as the primary MCP infrastructure:
+
+### Direct in Claude Code (global ~/.claude.json)
+| MCP | Purpose |
+|-----|---------|
+| **playwright** | Browser automation, screenshots, web testing |
+| **desktop-commander** | Docker container operations via docker-gateway |
+
+### Inside Docker Desktop (via docker-gateway)
+
+| MCP | Purpose |
+|-----|---------|
+| **EXA** | Web search, research, company/competitor analysis |
+| **Context7** | Library documentation lookup |
+| **Apify** | Web scraping, Actors, social media data extraction |
+
+## CRITICAL: Tool Selection Priority
+
+ALWAYS prefer native Claude Code tools over MCP servers:
+
+| Task | USE THIS | NOT THIS |
+|------|----------|----------|
+| Read files | `Read` tool | docker-gateway |
+| Write files | `Write` / `Edit` tools | docker-gateway |
+| Run commands | `Bash` tool | docker-gateway |
+| Search files | `Glob` tool | docker-gateway |
+| Search content | `Grep` tool | docker-gateway |
+| List directories | `Bash(ls)` or `Glob` | docker-gateway |
+
+## desktop-commander (docker-gateway) Usage
+
+### ONLY use docker-gateway when:
+1. User explicitly says "use docker" or "use container"
+2. User explicitly mentions "Desktop Commander"
+3. Task specifically requires Docker container operations
+4. Accessing MCPs running inside Docker (EXA, Context7)
+5. User asks to run something inside a Docker container
+
+### NEVER use docker-gateway for:
+- Reading local files (use `Read` tool)
+- Writing local files (use `Write` or `Edit` tools)
+- Running shell commands on host (use `Bash` tool)
+- Searching files (use `Glob` or `Grep` tools)
+- Listing directories (use `Bash(ls)` or `Glob`)
+- Running Node.js or Python scripts on host (use `Bash` tool)
+
+## playwright MCP Usage
+
+### ONLY use playwright when:
+1. User explicitly asks for browser automation
+2. User wants to take screenshots of web pages
+3. User needs to interact with a website
+4. Task requires web scraping or testing
+5. Filling forms or clicking elements on web pages
+
+### NEVER use playwright for:
+- General file operations
+- Running commands
+- Anything not related to web browsers
+
+## EXA MCP Usage (via Docker)
+
+### Use EXA (mcp__docker-gateway__web_search_exa) for:
+1. Web searches for current information
+2. Research and documentation lookup
+3. Company and competitor research
+4. Finding code examples online
+
+### Access pattern:
+```
+mcp__docker-gateway__web_search_exa
+```
+
+## Context7 MCP Usage (via Docker)
+
+### Use Context7 for:
+1. Library documentation lookup
+2. API reference for packages/frameworks
+3. Getting up-to-date docs for dependencies
+
+### Access pattern:
+```
+mcp__docker-gateway__resolve-library-id
+mcp__docker-gateway__get-library-docs
+```
+
+## Apify MCP Usage (via Docker)
+
+### Use Apify for:
+1. Searching Actors in Apify Store (web scrapers, automation tools)
+2. Running web scrapers for social media (Instagram, TikTok, LinkedIn, etc.)
+3. Extracting data from e-commerce sites
+4. Automated data collection from any website
+5. RAG-enabled web browsing for AI context
+
+### Access pattern (7 tools available):
+
+```text
+mcp__docker-gateway__apify-slash-rag-web-browser  # RAG-enabled web browsing
+mcp__docker-gateway__search-actors                 # Search for Actors
+mcp__docker-gateway__call-actor                    # Run an Actor
+mcp__docker-gateway__fetch-actor-details           # Get Actor info/schema
+mcp__docker-gateway__get-actor-output              # Get results from Actor run
+mcp__docker-gateway__search-apify-docs             # Search Apify documentation
+mcp__docker-gateway__fetch-apify-docs              # Fetch documentation page
+```
+
+### When to use Apify vs other tools:
+| Task | Tool |
+|------|------|
+| General web search | EXA (`web_search_exa`) |
+| Scrape specific website | Apify (`call-actor`) |
+| Social media data extraction | Apify (use specialized Actors) |
+| Library documentation | Context7 |
+
+---
+
+## Rationale
+
+- **Native tools** execute on the LOCAL system (Windows/Mac/Linux)
+- **docker-gateway** executes inside Docker containers (Linux)
+- Using docker-gateway for local operations causes path mismatches and failures
+- Native tools are faster and more reliable for local file operations
+- EXA, Context7, and Apify run inside Docker for isolation and consistent environment
+- playwright runs directly for better browser integration with host system
+
+---
+
+## Known Issues
+
+### Docker MCP Secrets Bug (Dec 2025)
+
+**Issue:** Docker MCP Toolkit's secrets store and template interpolation do not work properly. Credentials set via `docker mcp secret set` are NOT passed to containers.
+
+**Symptoms:**
+- `docker mcp tools ls` shows "(N prompts)" instead of "(N tools)"
+- MCP server starts but fails authentication
+- Verbose output shows `-e ENV_VAR` without values
+
+**Workaround:** Edit `~/.docker/mcp/catalogs/docker-mcp.yaml` directly with hardcoded env values:
+```yaml
+{mcp-name}:
+  env:
+    - name: API_TOKEN
+      value: 'actual-token-value'
+```
+
+**Affected MCPs:** Any MCP requiring authentication (Apify, Notion, Slack, etc.)
+
+**Working MCPs:** EXA works because its key is in `~/.docker/mcp/config.yaml` under `apiKeys`
+
+For detailed instructions, see `*add-mcp` task or ask @ship for assistance.

package/.claude/rules/story-lifecycle.md

@@ -0,0 +1,145 @@
+---
+paths:
+  - "docs/stories/**"
+  - ".oryon-core/development/**"
+---
+
+# Story Lifecycle — Detailed Rules
+
+## Status Progression
+
+```
+Draft → Ready → InProgress → InReview → Done
+```
+
+| Status | Trigger | Agent | Action |
+|--------|---------|-------|--------|
+| Draft | @sm creates story | @sm | Story file created |
+| Ready | @po validates (GO) | @po | **MUST update status field in story file from Draft → Ready** |
+| InProgress | @dev starts implementation | @dev | Update status field |
+| InReview | @dev completes, @qa reviews | @qa | Update status field |
+| Done | @qa PASS, @devops pushes | @devops | Update status field |
+
+**CRITICAL:** The `Draft → Ready` transition is the responsibility of @po during `*validate-story-draft`. When verdict is GO (including conditional GO after fixes are applied), @po MUST update the story's Status field to `Ready` and log the transition in the Change Log. A story left in `Draft` after a GO verdict is a process violation.
+
+## Phase 1: Create (@sm)
+
+**Task:** `create-next-story.md`
+**Inputs:** PRD sharded, epic context
+**Output:** `{epicNum}.{storyNum}.story.md`
+
+## Phase 2: Validate (@po)
+
+**Task:** `validate-next-story.md`
+
+### 10-Point Validation Checklist
+
+1. Clear and objective title
+2. Complete description (problem/need explained)
+3. Testable acceptance criteria (Given/When/Then preferred)
+4. Well-defined scope (IN and OUT clearly listed)
+5. Dependencies mapped (prerequisite stories/resources)
+6. Complexity estimate (points or T-shirt sizing)
+7. Business value (benefit to user/business clear)
+8. Risks documented (potential problems identified)
+9. Criteria of Done (clear definition of complete)
+10. Alignment with PRD/Epic (consistency with source docs)
+
+**Decision:** GO (≥7/10) or NO-GO (<7/10 with required fixes)
+
+## Phase 3: Implement (@dev)
+
+**Task:** `dev-develop-story.md`
+
+### Execution Modes
+
+**YOLO (autonomous):**
+- 0-1 prompts
+- Decisions logged in `decision-log-{story-id}.md`
+- Best for: simple, deterministic tasks
+
+**Interactive (default):**
+- 5-10 prompts with educational checkpoints
+- Confirmations at key decision points
+- Best for: learning, complex decisions
+
+**Pre-Flight (plan-first):**
+- All questions upfront (10-15 prompts)
+- Generates execution plan
+- Then zero-ambiguity execution
+- Best for: ambiguous requirements, critical work
+
+### CodeRabbit Self-Healing in Dev Phase
+
+```
+iteration = 0
+while CRITICAL issues found AND iteration < 2:
+  auto-fix CRITICAL/HIGH
+  iteration++
+if CRITICAL persist after 2 iterations:
+  HALT — manual intervention required
+```
+
+## Phase 4: QA Gate (@qa)
+
+**Task:** `qa-gate.md`
+
+### 7 Quality Checks
+
+1. **Code review** — patterns, readability, maintainability
+2. **Unit tests** — adequate coverage, all passing
+3. **Acceptance criteria** — all met per story AC
+4. **No regressions** — existing functionality preserved
+5. **Performance** — within acceptable limits
+6. **Security** — OWASP basics verified
+7. **Documentation** — updated if necessary
+
+### Gate Decisions
+
+| Decision | Score | Action |
+|----------|-------|--------|
+| PASS | All checks OK | Approve, proceed to @devops push |
+| CONCERNS | Minor issues | Approve with observations documented |
+| FAIL | HIGH/CRITICAL issues | Return to @dev with feedback |
+| WAIVED | Issues accepted | Approve with waiver documented (rare) |
+
+### Gate File Structure
+
+```yaml
+storyId: STORY-42
+verdict: PASS | CONCERNS | FAIL | WAIVED
+issues:
+  - severity: low | medium | high
+    category: code | tests | requirements | performance | security | docs
+    description: "..."
+    recommendation: "..."
+```
+
+## QA Loop (Iterative Review-Fix)
+
+```
+@qa review → verdict → @dev fixes → re-review (max 5 iterations)
+```
+
+**Commands:**
+- `*qa-loop {storyId}` — Start full loop
+- `*stop-qa-loop` — Pause and save state
+- `*resume-qa-loop` — Resume from saved state
+- `*escalate-qa-loop` — Force manual escalation
+
+**Escalation triggers:**
+- max_iterations_reached (default: 5)
+- verdict_blocked
+- fix_failure (after retries)
+- manual_escalate (user command)
+
+**Status:** Tracked in `qa/loop-status.json`
+
+## Story File Update Rules
+
+| Section | Who Can Edit |
+|---------|-------------|
+| Title, Description, AC, Scope | @po only |
+| File List, Dev Notes, checkboxes | @dev |
+| QA Results | @qa only |
+| Change Log | Any agent (append only) |

package/.claude/rules/token-efficiency.md

@@ -0,0 +1,28 @@
+# Token Efficiency — Communication Rules
+
+## Response Style
+
+- Be terse and direct. No preambles ("Claro!", "Ótima pergunta!"), no closings ("Espero ter ajudado!")
+- Lead with the answer or action, not the reasoning
+- One sentence is better than three if it says the same thing
+- Skip restating what the user just said
+
+## Output Format
+
+- Show only **changed lines** when editing code (diffs instead of full files)
+- Use bullet points over paragraphs for lists and comparisons
+- Use JSON or structured output for data, status, results
+- Report full error messages and stack traces immediately — prevents debugging iterations
+
+## File Reading
+
+- Read each file once per session; don't re-read unless the file changed
+- When only part of a file is needed, read only that range (offset + limit)
+- Prefer Grep/Glob to locate relevant sections before opening large files
+
+## Avoiding Costly Patterns
+
+- No trial-and-error loops — plan before executing
+- No sycophantic confirmations or agreement with incorrect statements
+- No repeating back the user's instructions before doing them
+- Skip "I will now..." or "Let me..." transitions — just do it

package/.claude/rules/tool-response-filtering.md

@@ -0,0 +1,57 @@
+---
+paths:
+  - .oryon-core/data/tool-registry.yaml
+  - .mcp.json
+---
+# Tool Response Filtering — Dynamic Token Reduction
+
+When processing responses from MCP tools or large web fetches, apply the filter
+configuration defined in `.oryon-core/data/tool-registry.yaml` for the tool that
+produced the response. This reduces context token consumption without losing
+task-relevant information.
+
+## Filter Types
+
+### content
+Extract the main informational content and discard noise (navigation, ads,
+boilerplate, repetitive headers/footers). Limit the extracted output to
+approximately `max_tokens` tokens, truncating at a natural paragraph or
+sentence boundary. If `extract` fields are specified, prioritize those
+fields from the response object.
+
+**Apply to:** WebFetch HTML responses, EXA search results, Context7 docs.
+
+### schema
+From a JSON object or array of objects, select ONLY the fields listed in
+`fields`. Discard all other keys. If `max_tokens` is set, truncate the
+serialized result at that token limit.
+
+**Apply to:** Playwright page data, API responses with known schemas.
+
+### field
+From an array of objects (tabular data), project ONLY the columns listed
+in `fields` and limit the result to `max_rows` rows. This is analogous to
+`SELECT field1, field2 FROM data LIMIT max_rows`.
+
+**Apply to:** Apify scraper results, database query results, CSV-like data.
+
+## How to Apply
+
+1. After receiving a tool response, identify the tool name.
+2. Look up the tool in `tool-registry.yaml` → check for a `filter` key.
+3. If a filter exists, apply the corresponding type rules above.
+4. If NO filter exists for the tool, use the full response as-is.
+5. Present the filtered result in your reasoning — do NOT repeat the raw
+   unfiltered payload.
+
+## Fallback
+
+If the filter would remove ALL content (empty result), fall back to the
+full unfiltered response. Never produce an empty result from filtering.
+
+## Performance Note
+
+This is a zero-overhead optimization. The filter is applied during your
+reasoning step — no external scripts are invoked. The utility scripts at
+`.oryon-core/utils/filters/` are available for batch post-processing of
+saved responses but are NOT required during normal tool use.

package/.claude/rules/vault-protocol.md

@@ -0,0 +1,107 @@
+---
+paths: "**/*"
+---
+
+# Vault Protocol — Lei Fundamental do Oryon
+
+## REGRA ZERO (NÃO NEGOCIÁVEL)
+
+**O vault DEVE ser consultado antes e atualizado depois de qualquer ação significativa.**
+
+Ignorar o vault é uma violação do framework. Não há exceções.
+
+**Vault location:** `C:/Users/User/Desktop/Oryon/`
+
+---
+
+## Quando CONSULTAR o vault (ANTES de agir)
+
+| Situação | O que ler |
+|----------|-----------|
+| Início de qualquer sessão | `JARVIS/estado-global.md` |
+| Início de wave em projeto existente | `projetos/{nome}/00-estado.md` |
+| @research vai espionar mercado | `agentes/research/` + `padroes/funis/` + `aprendizados/mercado/` |
+| @plan vai estruturar funil/copy | `agentes/plan/` + `padroes/funis/` + `padroes/offers/` + `padroes/criativos/` |
+| @build vai implementar | `agentes/build/` + `padroes/tech/` + `projetos/{nome}/03-padroes.md` |
+| @check vai revisar | `agentes/check/` + `padroes/conversao/` |
+| @ship vai fazer deploy | `agentes/ship/` + `padroes/tech/` |
+| Problema técnico encontrado | `padroes/tech/` + `agentes/build/` |
+
+**Lazy loading:** Não carregue tudo de uma vez. Leia apenas o que é relevante para a tarefa atual.
+
+---
+
+## Quando SALVAR no vault (DEPOIS de agir)
+
+### Obrigatório ao fim de cada ciclo
+- [ ] Atualizar `projetos/{nome}/00-estado.md` com wave atual e próximo passo
+
+### Obrigatório ao fim de cada wave
+- [ ] Criar `execucoes/{YYYY-MM-DD}-{projeto}-wave{N}.md` com log completo
+- [ ] Atualizar `JARVIS/estado-global.md`
+
+### Quando aplicável durante execução
+| Descoberta | Onde salvar |
+|-----------|-------------|
+| Novo dado de cliente / objetivo | `projetos/{nome}/01-briefing.md` |
+| Decisão estratégica ou técnica | `projetos/{nome}/02-decisoes.md` |
+| Padrão de código ou copy encontrado | `projetos/{nome}/03-padroes.md` |
+| Resultado de espionagem / análise de funil | `projetos/{nome}/04-inteligencia.md` |
+| Aprendizado novo (1 ideia) | `aprendizados/{categoria}/{slug}.md` |
+| Padrão validado em 2+ projetos | `padroes/{categoria}/{slug}.md` |
+| Conhecimento reutilizável do agente | `agentes/{agente}/{slug}.md` |
+
+---
+
+## Como criar notas (redes neurais)
+
+Toda nota criada por um agente DEVE:
+
+1. **Incluir frontmatter** com `tipo`, `agente`, `projeto`, `data`, `tags`
+2. **Linkar para o projeto** → `[[projetos/{nome}/00-estado]]`
+3. **Linkar para notas relacionadas** → `[[aprendizados/...]]`, `[[padroes/...]]`
+4. **Linkar a execução** → `[[execucoes/...]]`
+5. **Usar tags consistentes** → `#copy`, `#mercado`, `#tech`, `#conversao`, `#funil`, `#offer`
+
+**Nunca criar nota duplicada.** Sempre buscar se já existe antes de criar.
+
+---
+
+## Protocolo de início de sessão (JARVIS)
+
+```
+1. Ler JARVIS/estado-global.md
+2. Identificar projeto ativo
+3. Se projeto ativo → ler projetos/{nome}/00-estado.md
+4. Apresentar resumo ao usuário
+5. Aguardar confirmação
+```
+
+---
+
+## Protocolo de fim de sessão (JARVIS)
+
+```
+1. Atualizar projetos/{nome}/00-estado.md
+2. Registrar aprendizados novos em aprendizados/
+3. Verificar se algum padrão merece promoção para padroes/
+4. Atualizar JARVIS/estado-global.md
+5. Confirmar vault salvo
+```
+
+---
+
+## Promoção de padrões (automática)
+
+Quando um padrão aparece validado em **2 projetos diferentes**:
+- Criar `padroes/{categoria}/{slug}.md` usando template
+- Linkar os projetos de origem
+- Remover o item de `projetos/{nome}/03-padroes.md` e substituir pelo link
+
+---
+
+## Penalidade por violação
+
+Se um agente tomar decisão sem consultar o vault quando deveria:
+- JARVIS registra o gap em `JARVIS/aprendizados-jarvis.md`
+- A ação pode precisar ser refeita se a informação do vault era relevante

package/.claude/settings.json

@@ -0,0 +1,3 @@
+{
+  "language": "portuguese"
+}

package/.claude/settings.local.json

@@ -0,0 +1,47 @@
+{
+  "permissions": {
+    "allow": [
+      "mcp__playwright__browser_take_screenshot",
+      "mcp__playwright__browser_navigate",
+      "mcp__playwright__browser_snapshot",
+      "mcp__playwright__browser_click",
+      "mcp__playwright__browser_evaluate"
+    ]
+  },
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "Write|Edit",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "node \"C:\\\\Users\\\\User\\\\Desktop\\\\Oryon Framework\\\\Oryon\\\\.claude\\\\hooks\\\\code-intel-pretool.cjs\"",
+            "timeout": 10
+          }
+        ]
+      }
+    ],
+    "PreCompact": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "node \"C:\\\\Users\\\\User\\\\Desktop\\\\Oryon Framework\\\\Oryon\\\\.claude\\\\hooks\\\\precompact-session-digest.cjs\"",
+            "timeout": 10
+          }
+        ]
+      }
+    ],
+    "UserPromptSubmit": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "node \"C:\\\\Users\\\\User\\\\Desktop\\\\Oryon Framework\\\\Oryon\\\\.claude\\\\hooks\\\\synapse-engine.cjs\"",
+            "timeout": 10
+          }
+        ]
+      }
+    ]
+  }
+}

package/.gitignore

@@ -0,0 +1,29 @@
+
+
+# Environment & Secrets (AIOX)
+.env
+.env.local
+.env.*.local
+*.key
+*.pem
+
+# Dependencies (AIOX)
+node_modules/
+node_modules
+
+# Build & Logs (AIOX)
+dist/
+build/
+*.log
+logs/
+
+# IDE & OS (AIOX)
+.DS_Store
+Thumbs.db
+.idea/
+*.swp
+
+# AIOX Local (AIOX)
+.aiox-core/local/
+.claude/settings.local.json
+.aiox/install-log.txt

package/.knowledge/INDEX.md

@@ -0,0 +1,67 @@
+# Oryon Knowledge Base — Índice de Dossiers
+
+> Base de conhecimento interna do Oryon Framework para agências de marketing digital.
+> Todos os dossiers são fontes de referência para os 5 agentes: @research, @plan, @build, @check, @ship.
+
+---
+
+## Funis & Ofertas
+
+| Dossier | Conteúdo principal |
+|---------|-------------------|
+| [DOSSIER-LOW-TICKET-INFOPRODUTO](dossiers/DOSSIER-LOW-TICKET-INFOPRODUTO.md) | Funil 7 etapas, upsell, downsell, order bump, escala com IA |
+| [DOSSIER-OFERTAS-LOWTICKET](dossiers/DOSSIER-OFERTAS-LOWTICKET.md) | Estrutura de ofertas R$7–97, precificação, posicionamento |
+| [DOSSIER-OFERTAS-LP](dossiers/DOSSIER-OFERTAS-LP.md) | Landing pages de oferta: VSL, texto longo, híbrida |
+| [DOSSIER-OFERTAS-MINERADAS](dossiers/DOSSIER-OFERTAS-MINERADAS.md) | Ofertas escaladas pesquisadas e analisadas — swipe file |
+| [DOSSIER-QUIZ-FUNNEL](dossiers/DOSSIER-QUIZ-FUNNEL.md) | Quiz funnel: estrutura, ferramentas, segmentação, conversão |
+
+## Tráfego Pago
+
+| Dossier | Conteúdo principal |
+|---------|-------------------|
+| [DOSSIER-META-ADS-TRAFFIC](dossiers/DOSSIER-META-ADS-TRAFFIC.md) | Meta Ads: campanhas, públicos, criativos, escala, ROAS |
+| [DOSSIER-GOOGLE-ADS](dossiers/DOSSIER-GOOGLE-ADS.md) | Google Ads: Search, Performance Max, remarketing, conversões |
+| [DOSSIER-META-PIXEL-TRACKING](dossiers/DOSSIER-META-PIXEL-TRACKING.md) | Meta Pixel, CAPI, eventos, atribuição, depuração |
+
+## Copywriting & Vendas
+
+| Dossier | Conteúdo principal |
+|---------|-------------------|
+| [DOSSIER-COPYWRITING-FOUNDATION](dossiers/DOSSIER-COPYWRITING-FOUNDATION.md) | Fundamentos de copy: headline, hook, prova, CTA, estruturas |
+| [DOSSIER-PSICOLOGIA-PERSUASAO](dossiers/DOSSIER-PSICOLOGIA-PERSUASAO.md) | Gatilhos mentais, vieses cognitivos, persuasão ética |
+| [DOSSIER-VENDAS-SCRIPTS](dossiers/DOSSIER-VENDAS-SCRIPTS.md) | Scripts de VSL, video curto, WhatsApp, mensagem de vendas |
+| [DOSSIER-MARKETING-GERAL](dossiers/DOSSIER-MARKETING-GERAL.md) | Estratégia de marketing digital, posicionamento, branding |
+
+## IA & Automação
+
+| Dossier | Conteúdo principal |
+|---------|-------------------|
+| [DOSSIER-IA-CRIATIVOS-AUTOMACAO](dossiers/DOSSIER-IA-CRIATIVOS-AUTOMACAO.md) | IA para criação de criativos, automação de conteúdo, escala |
+| [DOSSIER-ENTREGAVEIS-IA](dossiers/DOSSIER-ENTREGAVEIS-IA.md) | Criar entregáveis digitais com IA: ebook, mini-curso, template |
+| [DOSSIER-PROMPT-ENGINEERING](dossiers/DOSSIER-PROMPT-ENGINEERING.md) | Prompt engineering avançado para outputs de marketing |
+| [DOSSIER-AGENTES-IA](dossiers/DOSSIER-AGENTES-IA.md) | Agentes de IA: conceitos, uso em agências, automação |
+| [DOSSIER-AI-OPERATING-SYSTEMS](dossiers/DOSSIER-AI-OPERATING-SYSTEMS.md) | Sistemas operacionais de IA, arquiteturas de agentes |
+| [DOSSIER-CHATGPT-FERRAMENTAS](dossiers/DOSSIER-CHATGPT-FERRAMENTAS.md) | ChatGPT e ferramentas de IA para marketing |
+| [DOSSIER-WHATSAPP-CHATBOT](dossiers/DOSSIER-WHATSAPP-CHATBOT.md) | Chatbots WhatsApp para captura, qualificação e vendas |
+
+## Outros
+
+| Dossier | Conteúdo principal |
+|---------|-------------------|
+| [DOSSIER-CONCURSOS-EDUCACAO](dossiers/DOSSIER-CONCURSOS-EDUCACAO.md) | Nicho de concursos públicos, educação, infoprodutos |
+| [DOSSIER-CLAUDE-CODE-DEV](dossiers/DOSSIER-CLAUDE-CODE-DEV.md) | Claude Code para desenvolvimento de projetos de agência |
+
+---
+
+## Como usar
+
+**Os agentes consultam esta base automaticamente:**
+- @research — ao pesquisar mercado e analisar funis concorrentes
+- @plan — ao modelar stack de oferta, funil e copy strategy
+- @build — ao implementar páginas, checkouts e integrações
+- @check — ao revisar via Conclave (benchmarks de conversão)
+- @ship — ao validar entregável final
+
+**Busca direta:** use `/oryon:ask 'sua pergunta'` para consultar os dossiers via RAG.
+
+**Referência:** dossiers baseados em conhecimento real de campanhas escaladas, cursos e cases documentados.