I Tried the "Colleague Skill" — Digital Cloning an Employee
The Colleague Skill project went viral in China. The premise: capture a real employee's communication style, problem-solving approach, and tech stack into an AI agent that can substitute for them.
How It Works
The system (built for Claude) uses three files to describe a person:
- Communication style — how they write, respond, express emotions
- Problem-solving approach — methodology, priorities, shortcuts
- Tech stack and expertise — tools, frameworks, domain knowledge
It assembles these from meta-information and actual chat histories. The result is a stylized prompt that attaches as a skill to Claude.
My Experience
I ran it. The concept is elegant — codify what makes a specific person effective.
But here's the reality: Claude as a coding agent is already excellent. The "colleague" framing adds narrative appeal, but the underlying capability is Claude's reasoning, not the personality capture.
Where It Gets Interesting
The project sparked a wave of "how do I prevent my employer from cloning me" discussions in China. Which reveals something important:
If your value is purely your communication style, you're replaceable. If your value is your judgment, network, and ability to navigate ambiguity — no prompt file captures that.
The Real Use Case
Not replacing people, but onboarding acceleration. A new team member can query "how would [senior engineer] approach this?" and get a reasonable approximation. It's a living documentation of decision patterns.
What's Missing
- Context awareness. The skill doesn't know current project state, blockers, or interpersonal dynamics.
- Evolution. People change their approach. Static skill files go stale.
- Error recovery. When the real employee would escalate or ask for help, the agent might confidently proceed down a wrong path.
My Take
Beautiful concept, limited practical value today. The real win is forcing teams to articulate why specific people are effective — which is valuable documentation regardless of AI.
More AI experiments, agent architectures, and production insights from inside a bank — follow my Telegram channel:
🚀 https://t.me/ai_tablet (Russian, technical)
Top comments (0)