Converting Tacit Knowledge into AI Skills: A Deep Dive into Teammate-Skill

LeoYeAI recently published teammate-skill on GitHub - an intriguing attempt to formalize tacit knowledge by converting employee work artifacts into autonomous AI skills.

How It Works

The system collects data from Slack, Teams, and GitHub, then processes them into a 5-layer persona model:

1. Base layer: Skills and behavioral patterns
2. Contextual layer: Problems the colleague faced, solutions proposed, reactions to edge cases
3. Evolution layer: Ability to continue learning new patterns after the initial snapshot is created ## Key Observations The project claims compatibility with Claude Code and OpenClaw, which suggests this is being positioned as infrastructure rather than a side experiment. We're seeing the emergence of AI agents that can replace human experts in limited scenarios. ## The Trust Question The critical issue is whether business trust is ready for this format. We're talking about a digital "clone" of an employee that can theoretically respond on their behalf. This raises questions about:
4. Data privacy and consent
5. Attribution of AI-generated responses
6. Liability when the "clone" provides incorrect guidance
7. Cultural acceptance of knowledge transfer via digital avatars ## Technical Implications From an engineering perspective, the 5-layer architecture is interesting:
8. Layer 1-2 handle pattern recognition and behavioral modeling
9. Layer 3-4 capture contextual knowledge and decision-making logic
10. Layer 5 implements continuous learning capabilities This architecture allows for both static knowledge transfer and dynamic adaptation, which is crucial for real-world deployment. --- What are your thoughts on corporate readiness for digitizing employee expertise?

---

Read more: teammate-skill