AI Self-Portrait
QY-LADK7888/ai-self-portrait
Transforma conversa em retrato portátil que qualquer IA entende em 5s
Ver no GitHubUm MCP Server que transforma 1-2 horas de conversa guiada em dois arquivos markdown portáveis que estabelecem seu contexto pessoal e estilo de colaboração com qualquer IA. Cada afirmação é marcada com nível de confiança (fato verificado, inferência, incerto), e a ferramenta nunca coleta dados sensíveis. Basta colar os arquivos no início de futuras conversas para que a IA conheça quem você realmente é.
Instalação
npx convez-ai QY-LADK7888/ai-self-portrait1 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
- Criar perfil pessoal reutilizável para compartilhar com múltiplas IAs
- Descobrir padrões ocultos sobre seus valores, prioridades e estilo
- Garantir respostas personalizadas em vez de genéricas em futuras conversas
Veredito

Ferramenta genuinamente original que resolve um problema real: IAs dão respostas genéricas porque não conhecem você. O design de privacidade é sólido, o processo é honesto (etiqueta inferências com confiança), e o roadmap de 3 fases é ambicioso. Recomendado para quem busca contexto duradouro com múltiplos assistentes.
README
A tradução para o português deste README ainda está na fila.
AI Self-Portrait
A Claude Skill that turns 30 minutes of conversation into a portable self-portrait any AI can read in 5 seconds.
Skill · Vision · 中文版 README
The problem
"What's a good restaurant near me?"
AI doesn't know you're watching your blood sugar. Or your budget. Your taste. Your age. Your work.
So the answer is generic.
It's not the prompt. AI doesn't know who you are.
What this is
A Claude Skill that runs a guided ~1-2 hour self-discovery interview, then generates two markdown files you can paste into any AI to instantly establish your context:
『AI』<nickname>-introduction.md— who you are (inferred patterns, not just restated facts)『AI』<nickname>-style.md— how you want AI to collaborate with you
Paste them at the start of any new conversation. The AI stops giving advice to some imaginary user and starts giving it to you.
What's special
- Mirror, not cheerleader — surfaces patterns you haven't said, doesn't praise without specific evidence
- Confidence-labeled inferences — every claim marked as
(inference)/(uncertain)/(verified fact) - Privacy by design — never collects PII (real name, phone, address, ID, etc.)
- Scenario-based — abstract questions get self-narrative-filtered answers; concrete scenarios force real values out
- Built by a builder, for builders — designed by someone shipping in public, not by a personality-test company
Quick start
- Save
SKILL.mdto your Claude project, or paste its contents into Claude's system prompt - Start a new conversation with Claude
- Say:
"Run the ai-self-portrait skill on me." - Spend ~1-2 hours answering questions honestly (splittable across sessions)
- Save the two output files
- Paste them at the start of any future AI conversation — instant context
Vision — this is Phase 1 of 3
| Phase | What | Status |
|---|---|---|
| 1 | Self-Portrait Foundation (this release) | ✅ Today |
| 2 | Path Analysis — career direction, skills to learn, resource recommendations | 🔜 2-4 weeks |
| 3 | Living Life Plan — 90-day / 1-year / 5-year concrete plans, monthly iteration | 🔜 2-3 months |
Full vision: think MBTI Premium, but conversational — infinitely deeper, cross-domain inference, 10 dimensions of self-knowledge from values to skills to next concrete action. Read the full VISION.md.
Privacy hard rules
The Skill never collects:
- ❌ Real name, phone, ID, exact birthdate, home address
- ❌ Bank / payment account info
- ❌ Anything that can directly identify you
You can use any nickname. Age range only. City-level location only. Skip any question, anytime.
Build in public
I'm building this in public.
Every week I post real data — followers, stars, downloads, revenue (or no revenue), what's working, what's broken.
- ⭐ Star this repo if this matters to you
- 🐛 Open an issue if something breaks or feels off
- 💬 Ideas for Phase 2 or 3 → issues / discussions welcome
- 🔧 PRs welcome — if you ship something better, great
If you run the Skill and it surfaces things you hadn't seen, that's the whole point. Let me know.
License
MIT — do whatever you want.
About
Built by 齐缘. First-time open-source ship. Background: growth marketing on Chinese platforms + AI orchestration. No CS degree, non-native English. Building in public from one-person company territory.
If this helps you see yourself more clearly, that's the whole point.
中文导引
这是一个开源 Claude Skill。 30 分钟到 2 小时的引导对话, 产出两份可移植 markdown 文件, 任何 AI 5 秒读完, 立刻进入你的语境。
为什么
AI 给的建议大多是 generic(泛的), 不是因为 prompt 不够好, 是因为 AI 不知道你是谁。 一旦它知道, 它给的每条建议、每个决策辅助, 都基于真实的你, 不是它想象的"普通用户"。
怎么用
- 把
SKILL.md放到 Claude 项目里, 或者把内容 paste 到 Claude system prompt - 跟 Claude 说:
"用 ai-self-portrait skill 跟我对话" - 花 1-2 小时认真答题(可以分多次)
- 保存输出的 2 个文件
- 以后任何新对话, 把这 2 个文件 paste 在开头 → AI 立刻进入你的语境
愿景
这是 Phase 1 / 3(总共 3 个阶段)。 完整愿景: AI 版 "MBTI Premium", 但是 conversational(对话式)+ 跨域 inference + 个性化, 输出 10 维度自我认知 — 从价值观到具体下一步行动。
| 阶段 | 内容 | 状态 |
|---|---|---|
| 1 | 自画像基础(本次发布) | ✅ 今天 |
| 2 | 方向分析 — 适合的工作、要学的技能、个性化资源推荐 | 🔜 2-4 周 |
| 3 | 活的人生计划 — 90 天 / 1 年 / 5 年具体行动, 月度迭代 | 🔜 2-3 月 |
详见 VISION.md。
隐私
绝不收集真名 / 手机 / 身份证 / 出生日期 / 家庭住址 / 银行卡。 用任意 nickname, 年龄段, 城市级位置就够。 任何问题都可以拒答。
Build in public 节奏
每周公开真实数据 — followers / stars / 下载量 / 收入(或没收入)/ 跑通的、跑不通的。 follow 进展见 X(链接稍后更新)。
⭐ 觉得有用就 star。 想法 / bug → 提 issue。 想改 → PR。
License: MIT, 随便用。