A custom AI Skill that analyzes photos, screenshots, posts, reels, stories, or conversations and generates strategic, psychology‑based conversation starters, follow‑ups, and relationship advice.
Note: The Skill file is named
namorar.skill(ornamorar.md) and its instructions are written in Portuguese because it was originally created by a Brazilian author.
Claude is multilingual — the Skill works perfectly in English, Spanish, French, German, Italian, Portuguese, and many other languages.
| File | Download Link |
|---|---|
namorar.skill (Recommended) |
⬇️ Download Latest |
namorar.md (Alternative) |
⬇️ Download Latest |
🔗 Repository: github.com/erikraft/flirting-skill
| Aspect | Details |
|---|---|
| Languages | Any language supported by the AI (English, Spanish, Portuguese, French, German, Italian, etc.) |
| Claude Models | Haiku 4.5, Haiku 4.5 Extended, Sonnet 3.5, Opus 3.5 |
| Claude Code | ✅ Works with Claude Code (via Skill import) |
| Browser Agentic | ✅ Works with Agentic browsers that support Skills |
| Local Models | ✅ Works with Ollama, Llama 3, Mistral, and other open‑source models |
| Platforms | Claude Web, Claude API, Claude Desktop, Claude Mobile |
| Attachments | Images, screenshots, PDFs (text extraction) — supports Image Analysis on compatible models |
| Platform | Type | Image Analysis | Skill Support | Tutorial |
|---|---|---|---|---|
| Claude Web | Cloud | ✅ Yes (via UI) | ✅ Native | Tutorial |
| Claude Desktop | Desktop App | ✅ Yes | ✅ Native | Tutorial |
| Claude API | API | ✅ Yes | ✅ Via prompt | Tutorial |
| Claude Code | CLI | ✅ Yes | ✅ Native | Tutorial |
| LibreChat | Self‑Hosted | ✅ Yes | ✅ Via prompt | Tutorial |
| Open WebUI | Self‑Hosted | ✅ Yes | ✅ Via prompt | Tutorial |
| Ollama + Open WebUI | Local | ✅ Yes (LLaVA, Moondream) | ✅ Via prompt | Tutorial |
| LM Studio | Local | ✅ Yes | ✅ Via prompt | Tutorial |
| GPT4All | Local | ✅ Yes | ✅ Via prompt | Tutorial |
| Jan.ai | Local | ✅ Yes | ✅ Via prompt | Tutorial |
| KoboldCPP | Local | ✅ Via prompt | Tutorial | |
| Agentic Browsers | Browser | ✅ Yes | ✅ Custom | Tutorial |
Important
For the best results:
- Use Claude Haiku 4.5 Extended for cloud‑based usage.
- For local usage, use Llama 3.2 Vision or LLaVA with Open WebUI.
- Enable Extended Thinking (Deep Reasoning) whenever available.
- For Image Analysis, choose a platform that supports vision models (Claude, GPT‑4 Vision, LLaVA, or Moondream).
-
Download the Skill file:
-
Go to Claude's Skills page:
🔗 https://claude.ai/customize/skills -
Click Create Skill.
-
Upload the downloaded file.
-
Save the Skill.
✅ The Skill is now available. Activate it by typing
/namorarin any chat.
-
Download the Skill file:
-
Open Claude Desktop.
-
Go to Settings → Skills.
-
Click Import Skill.
-
Upload the
.skillfile. -
Save and restart Claude Desktop.
-
Clone the repository:
git clone https://github.com/erikraft/flirting-skill.git cd flirting-skill -
Copy the Skill to your Claude Code skills directory:
cp namorar.skill ~/.claude/skills/ -
Or use it directly:
claude code --skill namorar.skill
-
Copy the contents of
namorar.mdinto your system prompt. -
Include it in your API request:
{ "system": "=== SKILL: /namorar - SISTEMA DE SEDUÇÃO E RELACIONAMENTOS (v9.0) === ...", "messages": [...] }
-
Download: ⬇️
namorar.md(Latest) -
In LibreChat, go to Settings → Custom Prompts.
-
Paste the entire content of
namorar.md. -
Save as System Prompt and enable it.
-
Use it with GPT‑4 Vision or Claude models.
-
Download: ⬇️
namorar.md(Latest) -
In Open WebUI, go to Admin Panel → System Prompts.
-
Create a New System Prompt named
namorar. -
Paste the content of
namorar.md. -
Save and enable it.
-
Use it with Ollama (Llama 3.2 Vision, LLaVA) or any compatible model.
-
Install Ollama:
curl -fsSL https://ollama.com/install.sh | sh -
Pull a Vision Model (recommended):
ollama pull llava ollama pull moondream
-
Install Open WebUI:
docker run -d -p 3000:8080 --name open-webui ghcr.io/open-webui/open-webui:main
-
Add the Skill:
- Open Open WebUI → Admin Panel → System Prompts
- Upload
namorar.mdcontent as a System Prompt
-
Use: Attach images and type
/namorar— the vision model will analyze the image and the Skill will generate the response.
-
Download LM Studio from lmstudio.ai.
-
Download a Vision Model (e.g., LLaVA, Moondream).
-
Go to Settings → System Prompts.
-
Paste the
namorar.mdcontent as a System Prompt. -
Enable the Skill and start chatting — the image analysis will be handled by the vision model.
-
Download GPT4All from gpt4all.io.
-
Go to Settings → Custom Instructions.
-
Paste the
namorar.mdcontent. -
Save and enable.
-
Use with any compatible model.
-
Download Jan from jan.ai.
-
Go to Settings → System Prompt.
-
Paste the
namorar.mdcontent. -
Save and enable.
-
Use with Jan Vision models for image analysis.
Agentic browsers (like Browser Use, Multi‑On, Jarvis):
-
Download: ⬇️
namorar.md(Latest) -
Open your Agentic Browser's Skill Manager.
-
Upload or paste the Skill content.
-
Enable the Skill.
-
The Agentic Browser will use it automatically when analyzing posts, profiles, and conversations.
| Step | Action |
|---|---|
| 1️⃣ | Type /namorar in the chat |
| 2️⃣ | Attach an image (photo, post, story, screenshot) — on vision‑enabled models |
| 3️⃣ | Or Describe the image if the model doesn't support vision |
| 4️⃣ | Specify your goal: "Start a conversation", "Reply to her story", "Continue the chat", "Get her to DM me" |
| 5️⃣ | The AI generates 3 strategic message options, each with a psychological justification |
- 📷 A photo (pose, outfit, background, expression)
- 💬 A conversation screenshot (what she said, the tone)
- 📸 An Instagram post (image + caption)
- 🎥 A Reel (what she's doing, the vibe)
- 📖 A Story (the context, the text, the setting)
- 🌐 Any online content you find interesting
The Skill is based on Bruno Kraus's method and behavioral psychology:
| Phase | Name | Mechanism | Neurochemical Target |
|---|---|---|---|
| 1 | Dopamine Trigger (Zeigarnik Effect) | An unfinished, intriguing statement that creates curiosity and makes her want to respond. | Dopamine — curiosity, obsession, anticipation |
| 2 | Oxytocin Connection (Cold Reading) | A deep observation about her personality with a playful contradiction, making her feel truly understood. | Oxytocin — bonding, trust, emotional connection |
| 3 | Desire Spark (Distorted Interpretation) | A playful, humorous twist on something she said or did, creating sexual tension without being explicit. | Testosterone + Dopamine — desire + anticipation |
| Image | Example Prompt |
|---|---|
 |
/namorar I found her on Threads. I want to start a conversation and convince her to DM me. |
Result: The AI analyzes the image, understands her vibe, posture, outfit, and setting — then generates 3 strategic replies tailored to the situation.
The Skill is not limited to first messages. You can keep using it throughout the conversation.
| Image | Example Prompt |
|---|---|
 |
/namorar She replied. What should I say now? |
Result: The AI reads the latest screenshot, understands the flow, and generates a follow‑up that maintains continuity.
- 🔍 Find a post, Reel, Story, or profile.
- 💬 Use
/namorarto generate an initial comment or first message. - ❤️ Receive a reply.
- 📷 Continue using
/namorarwith updated screenshots or descriptions. - 💌 When the conversation moves from public comments to DMs, keep attaching the latest screenshot and continue using
/namorar. - ✨ The AI keeps generating natural follow‑ups that flow with the conversation.
💡 Pro Tip: If the AI cannot analyze the image (e.g., Claude's safety filter or non‑vision model), simply describe the image in text. Example: "She's lying on a bed, black leggings, messy hair, smiling at the camera — the vibe is relaxed and flirty." The Skill works perfectly with text descriptions too.
| Issue | What happens | How to solve |
|---|---|---|
| Faces in images | Some AI models may refuse to analyze images with identifiable faces | Describe the image in text instead of attaching it |
| Explicit content | AI may refuse sexually explicit requests | Use suggestive, playful language instead of explicit terms |
| Underage subjects | AI will refuse — always ensure you're interacting with adults | The Skill is designed for adult relationships only |
| Non‑vision models | Cannot analyze images | Use the text description method |
The Skill itself does not cause refusals — the AI's built‑in safety system does. The Skill is designed to work around these limits by encouraging text descriptions when needed.
| Feature | Description |
|---|---|
| 3‑Phase System | Dopamine, Oxytocin, and Desire triggers |
| 9 Powerful Questions | Based on Bruno Kraus's "infalible questions" to create deep emotional connection |
| Dual Modes | Solteiro (Single) and Relacionamento (Relationship) — adapts the tone |
| Context‑Aware | Understands public posts, DMs, stories, and long conversations |
| Psychology‑Backed | Grounded in Zeigarnik Effect, Cold Reading, and Neurochemistry |
| No Generic Compliments | Never uses "linda", "gostosa", or other clichés |
| Image Analysis Ready | Works with vision models for automatic analysis |
| Cross‑Platform | Works on Claude, Claude Code, Agentic Browsers, and Local Models |
This project is licensed under the MIT License.
You are free to:
- ✅ Use it commercially and personally
- ✅ Modify and adapt it
- ✅ Distribute it
- ✅ Include it in your own projects
All you need to do is retain the original copyright notice and license text.
See the LICENSE file for the full license text.
Q: Does this Skill work in English?
A: Yes! AI models are multilingual. The Skill instructions are in Portuguese, but the AI understands them and generates replies in whatever language you're using.
Q: Can I use this with Claude Code or Agentic Browsers?
A: Yes! The Skill works with Claude Code, Browser Use, Multi‑On, and other agentic systems.
Q: Can I run this locally on my computer?
A: Yes! Use Open WebUI + Ollama with a vision model like LLaVA or Moondream. See the tutorial above.
Q: What if the AI refuses to analyze an image?
A: Simply describe the image in text instead. The Skill works just as well (or better) with descriptions.
Q: Does this Skill create explicit content?
A: No. It creates suggestive, playful, and confident messages — never explicit or vulgar.
Q: Can I use this for long‑term relationships?
A: Yes! The Skill has a dedicated Relacionamento (Relationship) mode designed to maintain attraction and connection in established partnerships.
Q: Which local model works best?
A: Llama 3.2 Vision, LLaVA, or Moondream for vision tasks. For text‑only, Llama 3.1 8B or Mistral 7B work well.
Q: How do I download the latest version?
A: Use the direct download links:
Feel free to fork this repository, improve the Skill, and submit a pull request.
If you have suggestions, open an issue or reach out.
- Bruno Kraus — Relationship coach and the inspiration behind the method.
- Original Creator — Brazilian author who developed and refined the Skill.
👑 We're in this together!
Use this Skill wisely, respectfully, and always with consent.