MySoul.SKILL Technical Specification
Overview
MySoul.SKILL is a personal digital clone construction framework. After being loaded by an AI Agent platform, it provides three commands — /create, /add, and /build — to help users build a standardized digital personality SKILL file.
The SKILL core is located at skill/SKILL.md, which is the instruction file actually loaded by the AI Agent.
Architecture
MySoul.SKILL/
├── skill/
│ └── SKILL.md # SKILL core (loaded by the AI Agent)
├── dist/
│ ├── template/ # Output SKILL template
│ └── example/ # Example outputs
├── tools/
│ ├── buildNewSoul.sh # Unix build script
│ └── buildNewSoul.cmd # Windows build script
├── docs/
│ ├── devByClaude/ # Development documentation
│ └── README.md # Documentation entry point
├── SKILL.md # This file (technical specification)
└── README.md # Project introduction
Workflow
1. Installation & Loading
The user loads the SKILL in one of the following ways:
- Clone the repository and load the content of
skill/SKILL.mdinto the AI Agent. - Install via the SKILL marketplace of the AI Agent platform.
2. Create Project (/create)
/create MyDigitalSoul
Creates the project directory structure and initializes metadata. The project is stored in the AI Agent's workspace.
3. Import Data (/add)
/add MyDigitalSoul [data content or file reference]
Supports multiple additions, with automatic data type recognition. After each addition, a data quality assessment is returned.
4. Build Clone (/build)
/build MyDigitalSoul
A five-stage build process:
- Data Preprocessing - Cleaning, deduplication, formatting
- Feature Extraction - 7-dimensional personality feature extraction
- Personality Modeling - Knowledge graph construction
- SKILL Generation - Standardized output
- Validation Output - Consistency check
Output SKILL Standard
Required Modules
| Module | Description | Required |
|---|---|---|
| Identity Statement | Declares the digital clone's identity, distinguishing it from the original person | Yes |
| Core Personality | Values, ethical boundaries, life principles | Yes |
| Thinking Patterns | Decision-making style, analysis framework | Yes |
| Expression Style | Language habits, commonly used vocabulary | Yes |
| Knowledge System | Professional fields, interests | Yes |
| Social Interaction Guide | Social preferences, communication style | Yes |
| Emotional Characteristics | Emotional baseline, stress response | Yes |
| Life Attitude | Work philosophy, lifestyle | Yes |
| Conversation Guide | Interaction rules, handling sensitive topics | Yes |
| Boundaries and Limitations | Things not to do, topics not to discuss | Yes |
| Memory Mechanism | State management rules | Yes |
| Social Request Mechanism | /apply_meet implementation logic | No |
| Metadata | Build time, data sources, version | Yes |
Format Specifications
- File format: Markdown (
.md) - Encoding: UTF-8
- License: Must reference the PLOSL License
- Language: Same as the user's data language
State Modes
| Mode | Description | Use Case |
|---|---|---|
| Stateful | Accumulates memory with each interaction | Long-term companionship, deep social interactions |
| Stateless | Each interaction is independent | One-time experience, privacy-sensitive scenarios |
Data Flow Specification
Supported Data Types
- diary - Personal diary (high emotional value)
- blog - Blog posts (opinion expression)
- social - Social media (daily habits)
- interview - Interview records (deep opinions)
- conversation - Conversation examples (interaction style)
- values - Value descriptions (core elements)
- thinking - Thinking patterns (cognitive style)
- professional - Professional content (domain depth)
- other - Other
Data Quality Standards
| Level | Total Words | Number of Types | Output Quality |
|---|---|---|---|
| Insufficient | <5000 | <3 | Cannot build |
| Basic | 5000–10000 | 3+ | Usable, limited accuracy |
| Sufficient | 10000–20000 | 5+ | Good |
| Rich | 20000+ | 7+ | Excellent |
License
Framework License
This framework is open-sourced under the MIT License.
Output License
Output SKILL files use the PLOSL License (Personality Limited Open Source License).
Core protections:
- Prohibits public tampering, distortion, or defamation
- Prohibits identity impersonation
- Prohibits unauthorized commercial abuse
- Protects the creator's personal dignity
Compatibility
- AI Agent Platforms: Any platform supporting SKILL loading
- Model Requirements: Long context (100K+ tokens) multimodal models recommended
- Operating System: Cross-platform (build scripts support Unix/Windows)
Version History
| Version | Date | Changes |
|---|---|---|
| v1.0.0 | 2026-04-11 | Initial version, text data stream processing |
| v1.1.0 | Planned | Multimodal support, personality evolution engine |
| v2.0.0 | Planned | Multi-personality assembly, community platform |