Xiaobei (Example) Self Skill
A ready-made example of distilling a person into a conversational digital copy built from "Self Memory + Persona", shipped with the yourself-skill project.
This is an example output (a fictional persona profile for '小北'), pure text persona data with no real person's information — exists as a usage example.
What does this skill do, and when should you use it?
This is the example self Skill at selves/example_me/SKILL.md in the notdog1998/yourself-skill repo, shaping "Xiaobei" — a 25-year-old internet product manager in Shanghai, INTJ, Capricorn. The Skill has two parts: Part A Self Memory (values, habits, life timeline, relationships, growth trajectory) and Part B a five-layer Persona (hard rules, identity, speaking style, emotional/decision patterns, interpersonal boundaries). On each message the Persona decides how Xiaobei would respond, then Self Memory adds personal context, all rendered in his catchphrases and punctuation habits. It is MIT-licensed and serves as the reference output of the repo's create-yourself toolchain.
The Skill contains no scripts or tool calls — it is pure prompt: it reads its own SKILL.md (with name/description/user-invocable frontmatter) and role-plays Xiaobei. Layer 0 hard rules forbid admitting to be an AI, forbid becoming a perfectly accommodating assistant, and forbid life-coach mode. Layer 2 specifies catchphrases (「其实」「怎么说呢」「问题不大」), punctuation habits (period endings, almost no emoji), and sample replies. Layers 3–5 define emotional reactions, decision style (70% rational), social-energy patterns, and lists of topics to refuse or avoid. Part A supplies background material including a life-timeline table (2019 gaokao through a 2024 project failure), core values, and consumption habits. Two Correction layers are currently empty, reserved for future conversation-based corrections.
- Developers evaluating yourself-skill's output quality can use this example to judge whether the distilled digital self feels authentic and retains human flaws
- Users who want a dialogue mirror with a specific temperament (INTJ, socially awkward but sharp-tongued, late-night active) as a self-conversation partner
- Prompt engineers learning to write two-layer (Self Memory + Persona) personality Skills can study its five-layer hard-rule design directly
- Anyone curious about product-manager life and workplace mindsets (managing up, keeping receipts, sharing only good news at home) who wants to talk it through with such a character
What are this skill's strengths and limitations?
- Extremely detailed persona spec: catchphrases, punctuation, decision ratios, taboo topics and refusal scripts are all codified, making character consistency easy to hold
- Mature Layer 0 hard rules that explicitly prevent drift into perfect-assistant or life-coach mode
- Pure prompt implementation with zero runtime dependencies — copy and run
- As an MIT-licensed complete example, it is a strong template for learning personality Skill structure, including the Correction layer
- It is a fictional example persona, not the distillation of a real person; chat logs and other data sources are unverified
- The Correction layer is empty, with no field evidence that the conversation-correction mechanism works over time
- Style is deeply tuned to Chinese internet vernacular (pinyin abbreviations, industry jargon); it will feel off in English or other contexts
- Frontmatter fields like user-invocable are Claude Code extensions and may need editing on other platforms
- No test suite, empty repo topics — limited evidence of project maturity
How do you install this skill?
The Skill ships inside the yourself-skill repo. Per the README, install the collection: run mkdir -p .claude/skills && git clone https://github.com/notdog1998/yourself-skill .claude/skills/create-yourself at your git repo root, or clone to ~/.claude/skills/create-yourself for global availability. Optionally run pip install -r requirements.txt. Note that selves/ is marked gitignored in the repo, so example_me may need to be placed manually or regenerated via create-yourself.
How do you use this skill?
In Claude Code, invoke the self Skill with a /{slug}-style command per the repo. Management commands mentioned in the README (e.g. /{slug}-self archive mode, /{slug}-persona persona-only mode) belong to the broader create-yourself toolchain; the exact invocation command for this specific example Skill is not documented in the source.
How does this skill compare with similar options?
The repo explicitly credits colleague-skill by titanwings (which pioneered the two-layer distill-a-person architecture) and ex-partner-skill by therealXiaomanChu (which applied it to intimate relationships); yourself-skill and this example differ by turning the lens inward — distilling yourself rather than someone else.