Build clone-style advisors with explicit thought hierarchy, clear boundaries, and reusable workflow assets.
clone-skills is an open-source repository of clone-building workflows extracted from a private monorepo. It is built for people who want more than tone imitation: the goal is to turn judgment style, source material, and workflow logic into reusable skills.
This repository currently publishes two clone-oriented workflows:
| Skill | Core Hierarchy | What It Helps You Build |
|---|---|---|
mind-clone-creator |
persona -> tool -> workflow -> decision | A clone based on your own experience, judgment, work habits, and expression style |
mind-clone-advisor |
corpus/compliance -> profile/prompt -> graph | A compliant advisor workflow built from public-source materials about another person |
The root SKILL.md is the router that decides which workflow to use.
Most clone projects stop at "write a prompt that sounds like someone." These workflows go further:
- They separate source-of-truth questions from prompt-writing questions, so the clone is grounded before it is styled.
- They make the clone structure explicit instead of hiding everything inside one large system prompt.
- They keep boundaries visible: what is supported, what still needs workflow design, and what must stop for compliance review.
- They give you reusable assets such as profiles, prompts, workflow artifacts, and evaluation outputs.
- They are easier to maintain, review, and extend than ad hoc persona setups.
The core idea of this repository is simple: a useful clone should expose how its thinking is layered.
- If you only want tone mimicry, a prompt may be enough.
- If you want reusable judgment, tool use, workflow behavior, or decision support, you need explicit hierarchy.
- If you want to inspect or refine the clone later, hierarchy is what makes that possible.
That is why these skills are built around thought hierarchy instead of a single monolithic persona prompt.
The two published workflows use different layer models. That matters because this repository treats clone-building as a stack, not a voice filter.
mind-clone-creator is built as an upward-expanding stack:
- Persona layer
- Tool layer
- Workflow layer
- Decision layer
The default deliverable starts from the persona layer first. Higher layers become meaningful only after a concrete work unit is defined, which keeps the clone grounded before it grows into workflow or decision support.
mind-clone-advisor is built as a compliance-first analysis stack:
- Corpus and compliance layer
- Profile and prompt layer
- Optional graph layer
When the graph layer is enabled, the concept hierarchy is modeled as:
- belief -> model -> topic
This gives you a way to represent not only what someone says, but how their ideas are organized.
- Build a digital twin that responds more like you in review, planning, and advisory contexts
- Turn your own work style into a clone that can later grow into a workflow-oriented work clone
- Build a public-materials-based advisor around a writer, investor, founder, or public thinker
- Produce structured clone assets instead of loose notes: profile files, prompts, evaluation outputs, and workflow artifacts
- Explore a person's thinking at multiple levels, from beliefs and models down to topics and evidence structures
- Decide whether you need a voice clone, a judgment clone, a workflow clone, or a decision-support clone
- Use
mind-clone-creatorwhen the source of truth is you: your answers, your materials, and your work style. - Use
mind-clone-advisorwhen the source of truth is a public corpus about another person and you need authorization, compliance, and source-quality checks.
If the source identity is still unclear, start from the root router and let it disambiguate.
In a skill-enabled environment, start with prompts like these:
I want to create my own digital twinI want to turn my experience into an AI advisorI want to build an advisor clone of someone from public materials
This repository is intentionally narrow in scope:
- Public self-clone creation
- Public-materials-based advisor creation
- Reusable prompts, scripts, references, and evaluation assets for those two flows
Local runtime artifacts, session logs, and environment-specific workspace data stay out of version control.
clone-skills/
├── SKILL.md
├── README.md
├── README.zh.md
├── mind-clone-advisor/
└── mind-clone-creator/
- SKILL.md: the root router for clone-related requests
- mind-clone-creator/README.md: self-clone workflow overview
- mind-clone-creator/README.zh.md: Chinese README for the self-clone workflow
- mind-clone-creator/SKILL.md: full skill contract for self-clone creation
- mind-clone-advisor/README.md: public-materials-based advisor workflow overview
- mind-clone-advisor/README.zh.md: Chinese README for the advisor workflow
- mind-clone-advisor/SKILL.md: full skill contract for the advisor workflow
- mind-clone-advisor/references/guide.md: longer Chinese guide for public-materials-based mind-clone work
- This repository does not claim a fully autonomous work-replacing agent by default.
mind-clone-advisorshould only be used with legitimate public-source workflows and proper authorization review.- Files under
mind-clone-advisor/registry/should be treated as templates, not production data. - Generated examples should be reviewed before publishing downstream forks.
- Prefer relative paths in local configs and generated artifacts.
- Keep local runtime outputs, logs, and session artifacts out of commits.
- Keep public docs aligned with the currently published modules.
MIT