Aller au contenu
TakoAPI
Retour aux skills

multi-user-workspace

par shun-dong

Web & Frontend Development
522Téléchargements rapportés par clawskills.sh
2Étoiles
1Vues
0 mention J'aimeConnecte-toi pour aimer ce skill
—Pas encore noté
Ta note

Résumé

Multi-user workspace management with sandbox permissions, user profiles, and relationship networks.

Description

Ce que fait ce skill

Configures per-user sessions in a shared AI workspace with sandbox isolation, user profiles, and relationship-aware context. Each user gets their own session key, an optional Docker container for file isolation, and a profile file the assistant reads at session start. Relationship files define what information can cross user boundaries.

Enables multi-user AI workspaces on a single installation by routing sessions to per-user Docker containers based on session key patterns, without requiring separate deployments.

Quand l'utiliser

  • Giving a contractor sandboxed access to an AI assistant without exposing other users' files
  • Running separate AI sessions for multiple team members on a shared server
  • Tracking relationships between users to control cross-user information sharing
  • Assigning different tool permissions to admins versus guests in one workspace
  • Maintaining personalized assistant context for each user via profile files

Exemple de workflow

ENTRÉE

User asks: start a session as bob

AGENT
  1. 1

    Reads session key via session_status to extract userId 'bob'

  2. 2

    Loads FRIENDS/bob.md to retrieve Bob's profile and preferences

  3. 3

    Scans RELATIONS/ for files matching 'bob' to load relationship context

  4. 4

    Applies Bob's sandbox configuration from openclaw.json with isolated /workspace directory

  5. 5

    Greets Bob by name with personalized context loaded

SORTIE

Session initialized with Bob's profile, relationships, and sandboxed workspace mounted at /workspace

Prérequis

  • Docker (required for sandbox mode)

Installation

clawhub install shun-dong/multi-user-workspace

Exécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.

Liens