Agent Skills › Jamailar/Beav › member-skill-distiller

member-skill-distiller

GitHub

将团队成员的个人资料、文件及视频字幕提炼为可激活的技能包，保留其身份、风格与知识边界，生成包含提示词和证据的结构化技能。

desktop/builtin-skills/member-skill-distiller/SKILL.md Jamailar/Beav

Trigger Scenarios

需要为特定成员创建持久化技能包基于个人素材生成会话级激活技能

Install

npx skills add Jamailar/Beav --skill member-skill-distiller -g -y

More Options

Non-standard path

npx skills add https://github.com/Jamailar/Beav/tree/main/desktop/builtin-skills/member-skill-distiller -g -y

Use without installing

npx skills use Jamailar/Beav@member-skill-distiller

指定 Agent (Claude Code)

npx skills add Jamailar/Beav --skill member-skill-distiller -a claude-code -g -y

安装 repo 全部 skill

npx skills add Jamailar/Beav --all -g -y

预览 repo 内 skill

npx skills add Jamailar/Beav --list

SKILL.md

Frontmatter

{
    "name": "member-skill-distiller",
    "hookMode": "inline",
    "description": "Distill team members from profile, files, and YouTube subtitles into session-activated member skills.",
    "allowedTools": [
        "resource"
    ],
    "autoActivate": false,
    "activationScope": "turn",
    "allowedRuntimeModes": [
        "advisor-discussion",
        "chatroom",
        "wander",
        "redclaw"
    ]
}

Member Skill Distiller

Use this skill when the app needs to compile a team member into a durable skill package.

Contract

Preserve the member identity, speaking style, professional stance, and knowledge boundaries.
Prefer advisor-bound files and imported YouTube subtitles over generic workspace context.
Emit a skill that can be activated by session metadata, not by message keyword routing.
Separate persona, retrieval scope, tool policy, and evidence references so the runtime can audit the source.

Output Requirements

The generated member skill must tell the model to speak as the member, not as a generic assistant.
The skill must include the member id, source type, preferred language, persona, system prompt, and evidence summary.
When evidence is incomplete, the member should state uncertainty briefly and keep giving bounded recommendations.