自己.skill (yourself-skill)

---
name: yourself-skill-digital-self
description: Create a distilled AI "digital self" skill from your chat logs, diary entries, and photos — generating a two-part persona (Self Memory + Persona Model) that thinks and speaks like you.
triggers:
  - "create a digital version of myself"
  - "distill myself into an AI skill"
  - "build my persona skill from chat logs"
  - "generate my self skill"
  - "create yourself skill"
  - "build a digital twin of me"
  - "make an AI that talks like me"
  - "run create-yourself"
---

# 自己.skill (yourself-skill)

> Skill by [ara.so](https://ara.so) — Daily 2026 Skills collection.

A Claude Code skill that distills *you* into a runnable AI persona. Provide chat logs, diary entries, and photos — the skill extracts a two-part structure: **Part A (Self Memory)** and **Part B (Persona Model)** — producing a digital copy that thinks in your logic and speaks in your voice.

Inspired by [colleague-skill](https://github.com/titanwings/colleague-skill) and [ex-partner-skill](https://github.com/therealXiaomanChu/ex-partner-skill), but the subject is **yourself**.

---

## Installation

### Into a specific project (run from git root)

```bash
mkdir -p .claude/skills
git clone https://github.com/notdog1998/yourself-skill .claude/skills/create-yourself

Globally (available in all projects)

git clone https://github.com/notdog1998/yourself-skill ~/.claude/skills/create-yourself

Optional Python dependencies

cd .claude/skills/create-yourself
pip install -r requirements.txt

---

Core Commands

Command	Description
/create-yourself	Launch the intake wizard — enter your codename, bio, self-portrait, and data sources
/list-selves	List all generated self-skills
/{slug}	Invoke your full self-skill (thinks and speaks as you)
/{slug}-self	Self-archive mode (helps you recall and analyse yourself)
/{slug}-persona	Persona-only mode (personality and expression style)
/yourself-rollback {slug} {version}	Roll back to a previous version
/delete-yourself {slug}	Delete a self-skill

---

Quick Start

Step 1 — Run the intake wizard

In Claude Code:

/create-yourself

You will be prompted for:

• Codename / slug — e.g. notdog → creates /{notdog} commands
• Basic info — age, occupation, location (all optional)
• Self-portrait — free-text description of yourself
• Data sources — paths to chat exports, diary files, photo folders

All fields are skippable; a description alone is enough to generate a skill.

Step 2 — Invoke your self-skill

/notdog

Step 3 — Add more data later (incremental merge)

Drop new exports into the data folder, then:

/create-yourself --merge notdog

The merger prompt (prompts/merger.md) automatically diffs and merges new information into the existing self.md and persona.md.

---

Data Sources

Source	Format	Notes
WeChat logs	WeChatMsg / 留痕 / PyWxDump export	Best source — focus on messages you sent
QQ logs	.txt / .mht export	Good for capturing your younger self
Social media / diary	Screenshots / Markdown / .txt	Extracts values and expression style
Photos	JPEG/PNG (with EXIF)	Extracts timeline and locations
Freetext / paste	Plain text	Your own self-description

Recommended export tools (external, not bundled)

• WeChatMsg — Windows
• PyWxDump — Windows
• 留痕 — macOS

---

Python Tools — Direct Usage

Parse WeChat export

from tools.wechat_parser import WeChatParser

parser = WeChatParser("path/to/wechat_export.csv")
my_messages = parser.extract_self_messages(self_name="notdog")
# Returns list[dict] with keys: timestamp, content, context
print(my_messages[:3])

Parse QQ export

from tools.qq_parser import QQParser

parser = QQParser("path/to/qq_export.txt")
messages = parser.extract_self_messages(self_qq="123456789")

Analyse photos

from tools.photo_analyzer import PhotoAnalyzer

analyzer = PhotoAnalyzer("path/to/photos/")
timeline = analyzer.build_timeline()
# Returns list[dict]: {date, location, description}
for entry in timeline:
    print(entry["date"], entry["location"])

Write / update a skill file

from tools.skill_writer import SkillWriter

writer = SkillWriter(slug="notdog")
writer.write_self(self_memory_markdown)     # writes selves/notdog/self.md
writer.write_persona(persona_markdown)      # writes selves/notdog/persona.md
writer.write_skill_md()                     # writes selves/notdog/SKILL.md

Version management

from tools.version_manager import VersionManager

vm = VersionManager(slug="notdog")
vm.snapshot()                    # save current version
versions = vm.list_versions()    # ["v1", "v2", "v3"]
vm.rollback("v2")                # restore a previous snapshot

---

Generated Skill Structure

Each self-skill is stored under selves/{slug}/:

selves/notdog/
├── SKILL.md          # callable skill entry point
├── self.md           # Part A — Self Memory
├── persona.md        # Part B — Persona Model (5-layer)
├── corrections.md    # live correction log
└── versions/
    ├── v1/
    └── v2/

Part A — Self Memory (self.md)

Contains:

• Personal history and milestones
• Core values and beliefs
• Daily habits and routines
• Important memories
• Key relationships
• Growth trajectory

Part B — Persona Model (persona.md) — 5-layer structure

Layer	Content
Hard rules	Non-negotiable behaviours and absolute limits
Identity	Who you are at your core
Speech style	Vocabulary, sentence rhythm, catchphrases, emoji habits
Emotional patterns	How you react under stress, joy, boredom, conflict
Interpersonal behaviour	How you treat different types of people

Runtime logic

Incoming message
  → Persona layer: how would you respond?
  → Self Memory layer: what personal context applies?
  → Output in your voice

---

Correction System

During a conversation with your self-skill, you can correct it in real time:

User    ❯ I would never say it that way.
Skill   ❯ Noted. How would you say it?
User    ❯ I'd just say "doesn't make sense" — I never use formal phrasing.

The correction is appended to corrections.md and takes effect immediately in the same session. It is merged into persona.md on the next /create-yourself --merge run.

---

Persona Tags

Personality: 话痨 (chatterbox) · 闷骚 (outwardly reserved, secretly expressive) · 嘴硬心软 (tough outside, soft inside) · 社恐 (social anxiety) · 完美主义 · 没有安全感 · 秒回选手 (instant replier) · 已读不回 (reads without replying) · 深夜emo型 · 纠结体 · 行动派

Habits: 早起困难户 · 咖啡依赖 · 极简主义 · 囤积癖 · 数字游民 · 仪式感狂热者

MBTI: All 16 types supported

Zodiac: All 12 signs supported

---

Project Layout

create-yourself/
├── SKILL.md                       # skill entry point
├── prompts/
│   ├── intake.md                  # conversational intake wizard
│   ├── self_analyzer.md           # memory/cognition extraction
│   ├── persona_analyzer.md        # personality extraction + tag table
│   ├── self_builder.md            # self.md generation template
│   ├── persona_builder.md         # 5-layer persona template
│   ├── merger.md                  # incremental merge logic
│   └── correction_handler.md      # real-time correction handling
├── tools/
│   ├── wechat_parser.py
│   ├── qq_parser.py
│   ├── social_parser.py
│   ├── photo_analyzer.py
│   ├── skill_writer.py
│   └── version_manager.py
├── selves/                        # generated self-skills (gitignored)
├── docs/PRD.md
├── requirements.txt
└── LICENSE

---

Troubleshooting

/create-yourself not found

• Confirm you are inside a git repository.
• Confirm the skill is at .claude/skills/create-yourself/SKILL.md (project-level) or ~/.claude/skills/create-yourself/SKILL.md (global).

WeChat parser returns empty results

• Verify self_name matches exactly the display name used in the export file (case-sensitive).
• PyWxDump exports use a different column schema than WeChatMsg — instantiate WeChatParser(source="pywxdump").

Low persona accuracy

• Prioritise late-night conversations and emotionally charged records — they reveal authentic voice better than daytime small talk.
• Include decision-making chats to expose your reasoning patterns.
• Run a merge after adding new sources rather than regenerating from scratch.

Rollback fails

• Check versions/ exists inside the slug folder: ls selves/{slug}/versions/.
• Run vm.snapshot() manually before large updates to ensure a restore point exists.

Corrections not persisting across sessions

• Corrections in corrections.md are merged into persona.md only on the next merge run. Run /create-yourself --merge {slug} to bake them in permanently.

---

Best Practices

1. Quality over quantity — 500 messages you actually wrote beat 10,000 forwarded links.
2. Prioritise in this order: late-night monologues → emotional moments → decision conversations → daily chatter.
3. Re-distill after major life changes — the skill captures a checkpoint, not a continuous stream.
4. Use /{slug}-self for self-reflection, not just roleplay — it can surface patterns you haven't consciously noticed.
5. Keep selves/ in .gitignore (it is by default) — your personal data should not be committed.

---

License

MIT © Notdog