A structured multi-round calibration workflow to design, audit, and refine AI agent personalities and operating rules.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install persona-calibration
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install persona-calibration using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Persona Calibration?
This skill provides a systematic framework for evolving an agent from a generic assistant into a specialized partner, advisor, or researcher. Instead of relying on vague prompts, it uses an interview-based approach to separate a core persona from its expression style and operating boundaries. By utilizing the Persona Calibration skill, developers can ensure their agents behave consistently according to specific philosophical beliefs and decision-making styles.
The skill moves beyond simple role-playing by extracting hidden dimensions behind requested personalities and validating them through structured questions. It bridges the gap between high-level user desires and the low-level configuration files that govern agent behavior in Openclaw Skills.
Persona Calibration Use Cases
- Designing a new persona for a fresh Openclaw Skills instance from scratch.
- Refining existing assistant personalities to be more proactive, specialized, or concise.
- Transitioning an agent's role from a generalist to a specific domain expert like a senior architect or researcher.
- Comparing and implementing persona frameworks derived from external industry examples or GitHub repositories.
- Conducting structured interviews to tune behavioral boundaries and communication preferences for multi-agent systems.
How Persona Calibration Works
- Baseline Scan: The skill analyzes existing persona dimensions from web sources or internal reference documents.
- Dimension Modeling: It establishes a framework covering identity, beliefs, and communication styles using predefined buckets.
- Calibration Rounds: It conducts a 5-round progressive interview covering core positioning, expression, scenarios, and boundaries.
- Question Design: It employs trade-off and boundary questions to reveal actual operating preferences rather than surface-level prose.
- Summarization: It reflects back interpretations of each round to the user to resolve ambiguities and tensions.
- Proposal Generation: It translates results into a technical plan for file modifications before any changes are applied.
- File Application: It updates SOUL.md, IDENTITY.md, USER.md, and MEMORY.md with approved changes.
- Trial Period: It initiates a feedback loop to monitor real-world performance and refine the feel after actual usage.
Persona Calibration Setup
To integrate this into your workflow, ensure your environment is configured for Openclaw Skills development and that the necessary reference templates are available.
# Navigate to your agent project directory
cd my-openclaw-agent
# Ensure the skill is accessible in your skills path
# This skill requires write access to your persona configuration files
The skill relies on specific reference documents such as references/persona-dimensions.md and references/questionnaire-template.md to provide the underlying structure for the calibration rounds.
Persona Calibration Data Schema & Taxonomy
The Persona Calibration skill manages information across four primary Markdown files to ensure architectural clarity within Openclaw Skills.
| File | Purpose | Data Type |
|---|---|---|
SOUL.md |
Core personality, behavioral rules, scenario guidance, and absolute don'ts. | Markdown |
IDENTITY.md |
High-level role summary, creature type, or vibe description. | Markdown |
USER.md |
Specific preferences related to the human operator rather than the agent. | Markdown |
MEMORY.md |
Compact long-term summary of calibration results and significant persona updates. | Markdown |
Persona Calibration Advanced Features
- Progressive interview rounds designed to prevent user fatigue while gathering high-granularity behavioral data.
- Dimension extraction from industry examples to ground the persona in proven frameworks used across Openclaw Skills.
- Conflict detection to highlight tensions between competing values, such as the trade-off between conciseness and thoroughness.
- Multi-file mapping strategy that separates ephemeral communication styles from durable operating rules.
- Support for conditional switching, allowing personas to remain flexible and adapt their behavior based on specific task contexts.
SKILL.md
Loading
Related Openclaw Skills
Legal Practice Operations
A specialized AI agent skill for managing legal practice operations, case lifecycles, and strict court deadlines.
Hospitality Operations Agent
A comprehensive AI agent designed to manage end-to-end hotel operations, from guest reservations and housekeeping to maintenance and revenue optimization.
Nonprofit & NGO Operations
A comprehensive operational framework designed to automate and manage the complex workflows of nonprofit organizations and NGOs.
Memory Architecture
A professional-grade structured memory framework that uses hierarchical file storage and proactive hygiene to overcome LLM context window limitations.
Agent Browser Core
A production-ready web automation skill that enables AI agents to interact with the web using a Rust-based CLI, structured snapshots, and reference-based commands.
Browser-use
A high-performance CLI tool for persistent browser automation, web testing, and rapid data extraction with sub-50ms latency.
Featured*
