Summary

This project addresses the loss of critical team knowledge when members depart by creating AI "Skills" that emulate their work style and persona. It targets engineers and teams seeking to preserve and transfer tacit knowledge, enabling AI agents to operate with the context and expertise of specific individuals, thereby mitigating knowledge gaps.

How It Works

The system employs a dual-track AI analysis. Users provide a textual description and optionally ingest data from sources like Slack, GitHub, or documents. One track distills a "Work Skill" encompassing systems, standards, and workflows, while the other generates a "Persona" based on a 5-layer model of personality and communication. The output is a SKILL.md file, which supports continuous evolution through data appending and auto-versioning.

Quick Start & Requirements

• Primary Install:
  • OpenClaw/MyClaw.ai: openclaw skills install create-teammate
  • Claude Code: git clone ... into .claude/skills/ or ~/.claude/skills/
  • Other AgentSkills: Clone into the agent's skill directory.
• Prerequisites: Python 3.9+. Optional dependencies (Slack/GitHub auto-collectors) require API tokens; detailed setup is in INSTALL.md.
• Links: INSTALL.md, ClawHub, OpenClaw.

Highlighted Details

• Automated data ingestion from Slack, GitHub, email, Notion, Confluence, and more.
• Generates distinct Work Skill (technical capabilities) and 5-layer Persona (behavioral traits).
• Supports continuous skill evolution via data appending, with automatic versioning and rollback.
• Features a 3-layer quality assurance pipeline: pre-preview gate, post-creation smoke tests, and PII privacy scans.
• Enables side-by-side comparison of skills, including scenario simulation and decision debates.
• Skills are exportable as portable .tar.gz packages for easy sharing and import.

Maintenance & Community

The project is powered by MyClaw.ai and built for OpenClaw and Claude Code. Specific community channels or active maintainer details are not detailed in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Designed for native compatibility with OpenClaw and Claude Code agent frameworks. Permissive for commercial use under MIT terms.

Limitations & Caveats

The README does not explicitly detail limitations such as alpha status or known bugs. The focus is on feature richness and setup. Handling of Personally Identifiable Information (PII) is addressed via a privacy scan, with raw data excluded by default during export, suggesting careful consideration is needed for sensitive data.