Growth Engine
sofyanjamil/growth-engine
Pesquisa + 180 posts IG + SEO 5-pilares, gerados automaticamente do seu código
Ver no GitHubFerramenta end-to-end que gera automaticamente pesquisa de mercado, campanha Instagram de 30 dias com 180 pieces prontos para publicação, e conteúdo SEO em 5 pilares a partir da análise do seu repositório. Funciona como skill no Claude Code ou como agente portável em markdown para qualquer LLM.
Instalação
npx convez-ai sofyanjamil/growth-engine1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Startups precisam validar produto-mercado com pesquisa estruturada e go-to-market em paralelo
- Produtos técnicos geram conteúdo SEO em massa para capturar tráfego de comparações e alternativas
- Equipes pequenas automatizam calendário de redes sociais com conteúdo alinhado ao posicionamento
Veredito

Growth Engine é uma solução ambiciosa e bem-estruturada, mas seu sucesso depende fortemente da qualidade dos insights do modelo sobre seu codebase específico. O fluxo é portátil e o escopo (pesquisa + conteúdo + SEO) é generoso, embora exija refinamento manual em produção; adequado para produtos B2B técnicos e startups que dominam um domínio bem-definido.
README
A tradução para o português deste README ainda está na fila.
growth-engine
End-to-end growth pipeline for any codebase. Market research → 30-day Instagram campaign → 5-pillar SEO — produced from your own repo, in one command.
60-second quickstart
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/sofyanjamil/growth-engine/main/scripts/install.sh | bash
Windows (PowerShell):
iwr -useb https://raw.githubusercontent.com/sofyanjamil/growth-engine/main/scripts/install.ps1 | iex
Then, in any Claude Code session inside your project:
/growth
That's it. It will scan your codebase, ask 5–10 adaptive questions (or none if you pass --auto), and write a complete growth/ folder with research, 180 Instagram pieces, and 30 SEO pages.
Not using Claude Code? See docs/using-with-any-agent.md — the entire workflow runs on any LLM via
AGENT.md.
What you get
After one /growth run, your project has a growth/ folder:
growth/
├── 01-research/
│ ├── research.md # 8-framework market validation, 3-5k words
│ ├── research.html # styled report
│ └── verdict.json # PROCEED / PIVOT / KILL + 1-10 score
│
├── 02-content/
│ ├── calendar.json # 30 days × 6 pieces = 180 IG pieces
│ ├── calendar.md # human-readable schedule
│ ├── dashboard.html # visual overview
│ └── per-day/day-NN.json # full content per day (stories, carousel, post, reel)
│
├── 03-seo/
│ ├── pages/
│ │ ├── blog/ # 3-15 long-form articles
│ │ ├── vs/ # 2-10 head-to-head comparison pages
│ │ ├── alternatives/ # 2-10 alternatives-to pages
│ │ ├── integrations/ # 2-10 integration pages
│ │ └── glossary/ # 5-60 term definitions
│ ├── llms.txt # AI-crawler index (ChatGPT, Perplexity, Claude)
│ ├── llms-full.txt # full page corpus for AI ingestion
│ ├── sitemap.xml # Google + Bing
│ ├── robots.txt
│ └── WIRE-UP.md # framework-specific install instructions
│
└── references/ # ICP, positioning, voice, competitors, keywords
See docs/examples.md for a structural walkthrough of a full run.
The 3 phases
1. Market research — /growth-market
Runs 8 frameworks: Lean Startup hypothesis validation • Business Model Canvas + Porter's 5 + SWOT • Unit economics + CAC/LTV • Competitive analysis • PMF signals + retention curves • Real-world data sources • AI landscape • Funding strategy.
Outputs a single PROCEED / PIVOT / KILL verdict with a viability score and the top 5 risks + opportunities.
2. 30-day Instagram campaign — /growth-content
180 publish-ready pieces (3 stories + 1 carousel + 1 post + 1 reel × 30 days), value-first 60/30/10 mix (trust / engagement / product), themed weekdays, scored against an internal virality pattern library.
Every piece includes hashtags (volume-mixed), SEO-optimized captions, image prompts (Gemini 2.5 / GPT Image 2.0 format), shot lists for reels, and sticker prompts for stories.
3. 5-pillar SEO + AI-SEO — /growth-seo
14–105 pages across /blog, /vs, /alternatives, /integrations, /glossary — each with schema.org JSON-LD, FAQ blocks, citation magnets, og_image briefs, and internal-link graph. Plus llms.txt for AI crawler optimization.
Optional wire-up step (/growth-wire) auto-detects your framework (Next.js, Astro, SvelteKit, Vue, plain HTML) and applies idempotent edits to ship the SEO pages live.
Commands
| Command | What it does |
|---|---|
/growth | Orchestrator — runs all 3 phases end-to-end |
/growth-market | Market validation only |
/growth-content | 30-day IG campaign only |
/growth-seo | 5-pillar SEO only |
/growth-wire | Wire generated SEO into your codebase (idempotent) |
Flags
| Flag | Default | What it does |
|---|---|---|
--auto | off | Fully unattended — skip all Q&A, log inferred assumptions |
--days N | 30 | Campaign length in days |
--depth lean|standard|deep | standard | SEO volume: 14 / 30 / 105 pages |
--lang he,ar,en | en | Output language(s) — multi-lang supported |
--md-only | off | Skip HTML rendering, markdown only |
--no-wire | off | Skip the auto-wire prompt |
--url <url> | auto | Override live-URL detection |
--brief "..." | empty | Supplement codebase scan with a product description |
--target-users N | 2500 | Conversion target for content/SEO sizing |
--target-impressions N | 1000000 | Top-of-funnel target |
Install options
Option A — Claude Code plugin (recommended)
# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/sofyanjamil/growth-engine/main/scripts/install.sh | bash
# Windows PowerShell
iwr -useb https://raw.githubusercontent.com/sofyanjamil/growth-engine/main/scripts/install.ps1 | iex
This clones into ~/.claude/plugins/growth-engine/ and the /growth* commands appear automatically next session.
Option B — Manual clone
git clone https://github.com/sofyanjamil/growth-engine ~/.claude/plugins/growth-engine
Option C — Vendor into your repo (version-locked)
git clone https://github.com/sofyanjamil/growth-engine tools/growth-engine
# then symlink into the plugin dir, or commit tools/growth-engine/ for team-wide use
Option D — Other AI coding agents (Cursor, Codex, Cline, etc.)
git clone https://github.com/sofyanjamil/growth-engine
Then point your agent at AGENT.md — a single self-contained prompt that walks any LLM through all 3 phases using only file reads + web search. See per-agent guides:
Architecture
Free APIs only — uses WebSearch + WebFetch for competitor teardown, Google autosuggest, People-Also-Ask mining, and Reddit/Quora question mining. No OpenAI, Anthropic, Stripe, or SERP API keys required.
Model split — strategy work runs on the parent model (Opus if your session is Opus); writing-heavy fan-out drops to Sonnet via explicit model: sonnet on subagents:
| Work | Model | Why |
|---|---|---|
| 8-framework synthesis + verdict | Parent (Opus) | high-cognition |
| Orchestration + Socratic Q&A | Parent | high-cognition |
| Voice ingestion + calendar planning | Parent | strategy |
| Writing 180 IG pieces | Sonnet × 4 parallel | volume |
| Keyword consolidation + page-brief authoring | Parent | strategy |
| Writing 14–105 SEO pages | Sonnet × 5 parallel (one per pillar) | volume |
| Wire-up file edits | Parent | low-volume |
Result: a /growth run uses Opus only for the ~30k high-cognition output tokens and offloads ~200k writing tokens to Sonnet. Rough cost: 3–5× cheaper than Opus-only, with no observable quality drop on the writing layer.
What's inside
growth-engine/
├── .claude-plugin/plugin.json
├── skills/ # 3 skills (market-validate, content-month, seo-engine)
├── commands/ # 5 slash commands
├── agents/ # 2 parallel writers (Sonnet)
├── references/ # 22 reference docs (frameworks, virality, SEO patterns, schemas)
├── docs/ # per-agent install guides
├── scripts/ # install.sh + install.ps1
├── AGENT.md # single-file prompt for non-CC agents
└── README.md (this file)
Roadmap
- v0.2 — MCP server adapter (parity for Claude Desktop / Cursor / Windsurf via tool calls)
- v0.3 — Claude Code plugin marketplace listing
- v0.4 — Optional CLI wrapper (
npx growth-engine) for fully agent-agnostic runs - v0.5 — TikTok + LinkedIn campaign skills
- v0.6 — Localization beyond he / ar / en (es, fr, de, pt-BR)
See CHANGELOG.md for release history.
Contributing
Issues + PRs welcome. The plugin is markdown-only — no build step, no tests, no CI. Edit the relevant SKILL.md / command.md / agent.md and the change ships next session.
When adding a new skill or command:
- Follow the patterns in
skills/market-validate/SKILL.md— frontmatter + flow + output schema. - Reference shared content in
references/rather than duplicating. - Update this README's commands/flags tables.
- Bump version in
.claude-plugin/plugin.jsonand add a CHANGELOG entry.
License
MIT © 2026 Sofya — see LICENSE.