Cyber Colleague

What It Solves

A lot of team knowledge sits in Feishu, DingTalk, and Slack threads, email, docs, and screenshots. The useful parts are often a coworker’s review habits, system boundaries, and communication style. This skill turns that material into a reusable, updatable colleague profile instead of a shallow summary.

How It Works

It starts by confirming a name or alias, a short role summary, and optional tags such as personality, communication style, or team impression. Input can come from auto-collection, file upload, pasted text, or manual info only. It can parse Feishu JSON, .eml, .mbox, PDF, screenshots, Markdown, and TXT, and for Feishu links it chooses between browser mode and MCP mode.

Analysis runs along two tracks:

• Work capability: responsibility scope, system boundaries, stack, review focus, and output preferences
• Communication and style: expression, decision habits, collaboration style, and common tone

Before writing files, it previews the Work summary, Persona summary, and remaining uncertainties. After confirmation, it writes to ./colleagues/{slug}/ with meta.json, work.md, persona.md, and a merged SKILL.md. Existing profiles are updated incrementally with version history preserved.

Limits To Keep In Mind

It is best used when enough evidence exists to support the conclusion. It should not invent behavior from thin material. When using third-party collection, credentials should come from environment variables, and real chats, docs, or credentials should stay out of release archives. Missing permissions, dependencies, or insufficient evidence should be reported explicitly rather than masked as success.