"Colleague.Skill" Rises to Trending, Departed Colleagues Have Been Refined

What’s happening

A new open‑source project called Colleague.skill has exploded on GitHub, reportedly drawing some 6.7k stars in just a few days as developers use chat logs and documents to build AI “digital colleagues.” The idea is simple and unsettling: feed an AI a departed co‑worker’s WeChat (微信), Feishu (飞书), DingTalk (钉钉) or email exports and it will synthesize a persona that mimics their tone, decision patterns and even their habit of deflecting blame. It has been reported that contributors say the tool can reproduce 1:1 conversational quirks — the long memos, the curt replies, the technical nitpicks.

How it works and why it spread

Colleague.skill stitches together exported messages, meeting notes and doc histories using off‑the‑shelf chat‑dumping tools and large language models. Reportedly, the project’s design layers strict “hard” rules (coding standards, required comments) with style, decision heuristics and interpersonal behavior to produce a reusable “skill” that answers questions in the departed person’s voice. The repo has already inspired offshoots — ex‑partner.skill, boss.skill and even parent.skill — reflecting a broader push to turn institutional knowledge into packaged digital assets.

Industry and geopolitical context

The trend sits alongside industry moves to scale so‑called digital employees: NVIDIA (英伟达) recently floated AI‑compute token ideas and senior figures in Silicon Valley have discussed assistants that remember whole lives. It has been reported that companies are positioning AI as an efficiency multiplier while nations race for AI chips and talent under export controls and other trade frictions. In short: Western firms fret about existential AI risks; many Chinese developers are applying the tech to very pragmatic workplace problems — like who should fix the legacy code.

Legal and ethical fault lines

That pragmatism raises thorny questions. Who owns a departed employee’s chat logs and annotations — the person, the company, or the AI that ingests them? If an intern’s three years of messages can bootstrap an “80%” senior architect, where does experience value go? And if a digital twin continues to answer pings at 3 a.m., did the person really leave, or have they been turned into perpetual labor? Policymakers and HR teams may have to catch up fast — before cyber immortality becomes a routine HR process.