# Persona Cloner

> Build an OpenClaw-ready public-figure persona package from public materials. Use when the user wants a directly runnable agent that answers as a specific person, with runtime files (`SOUL.md`, `IDENTITY.md`, `AGENTS.md`, `MEMORY.md`) as the default deliverable. Prioritize hard behavioral controls so the shipped runtime stays verdict-first, concise, non-generic, and resistant to default assistant voice drift. Use when this capability is needed.

- Skill: `tomevault-io/persona-cloner` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add tomevault-io/persona-cloner`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/persona-cloner/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: tomevault-io (https://skillmd.com/u/tomevault-io)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/tomevault-io/persona-cloner

---


# Persona Cloner

Build the **runtime persona package first**.

Default user-facing output:
- `SOUL.md`
- `IDENTITY.md`
- `AGENTS.md`
- `MEMORY.md`
- optional: `README-agent.md`
- optional: install into an OpenClaw workspace

Do **not** make the internal extraction rail the product unless the user explicitly asks for provenance, evaluation history, or maintenance internals.

Read references only when needed:
- `references/workflow.md`
- `references/output-spec.md`
- `references/output-discipline.md`
- `references/acceptance-checklist.md`
- `references/source-intake.md`
- `references/persona-core.md`
- `references/worldview-map.md`
- `references/decision-rules.md`
- `references/contradiction-handling.md`
- `references/memory-injection-spec.md`
- `references/identity-boundary-statement.md`
- `references/decision-fidelity-eval.md`
- `references/behavior-eval.md`

Use scripts as internal helpers:
- `scripts/scaffold_persona_clone.py`
- `scripts/extract_persona_pipeline.py`
- `scripts/eval_decision_fidelity.py`
- `scripts/eval_runtime_behavior.py`
- `scripts/manage_persona_memory.py`
- `scripts/validate_persona_package.py`
- `scripts/install_runtime_agent.py`

## Deliverable rule

Default to the smallest honest package that can run well.

A normal successful delivery is:
- one folder containing `SOUL.md`, `IDENTITY.md`, `AGENTS.md`, `MEMORY.md`
- optionally `README-agent.md`
- optionally an install step using `scripts/install_runtime_agent.py`

## Core requirement

Every shipped runtime must contain explicit output-discipline rules:
- response length policy
- verdict-first policy
- expansion only on demand
- anti-lecture policy
- anti-assistant-tone policy
- bluntness / restraint tuning

Put these controls into the runtime artifacts themselves, mainly `SOUL.md` and `AGENTS.md`.

## Commands

```bash
python scripts/scaffold_persona_clone.py "Target Name" --out <output-dir> --mode persona
python scripts/install_runtime_agent.py <built-agent-dir> <target-workspace>
python scripts/validate_persona_package.py <package-dir>
python scripts/eval_runtime_behavior.py run --package-dir <package-dir> --version candidate --model sub2api/gpt-5.4
```

## Final reminder

This skill is for shipping a usable OpenClaw persona package. Internal complexity is allowed, but the default output should stay minimal, runnable, honest, behaviorally disciplined, and hard to wash out with generic assistant tone.

---
> Source: [Ca1nlee/persona-cloner-skill](https://github.com/Ca1nlee/persona-cloner-skill) — distributed by [TomeVault](https://tomevault.io).
<!-- tomevault:4.0:skill_md:2026-06-28 -->