colleague-skill

colleague.skill is a tool for capturing what a departing colleague, intern, mentor, or partner knew and how they worked, and turning that into an AI Skill you can keep using after they're gone. The pitch in the README is to turn cold goodbyes into warm Skills: instead of losing the context, voice, and habits of someone who has moved on, you distill them into something an AI assistant can act out on your behalf.

You feed it source materials from places like Feishu, DingTalk, Slack, WeChat, email, PDFs, Markdown notes, screenshots, or directly pasted text, plus your own subjective description of the person, such as their level, personality tags, and corporate-culture style. From that input it generates a Skill with two halves: a Work Skill that captures their systems knowledge, technical standards, and workflows, and a Persona that captures how they speak and make decisions across five layers from hard rules to interpersonal style. When a task comes in, the Persona decides the attitude and the Work Skill carries out the actual job, so the response sounds and acts like that person.

It is meant to plug into AI coding assistants. The README describes installing it under .claude/skills/ for Claude Code, or under an OpenClaw skills directory, then invoking commands like /create-colleague to build a new one, /list-colleagues to see what you have, and /{slug} to talk to a specific colleague. It is built in Python (3.9+ per the badge) and is open-source under the MIT license, with the project evolving into a broader effort called dot-skill described in its roadmap. The full README is longer than what was provided.