Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Multi User Workspace
v1.0.0Multi-user workspace management with sandbox permissions, user profiles, and relationship networks.
⭐ 2· 712·4 current·4 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The SKILL.md explains managing per-user session keys, reading per-user FRIENDS/{userId}.md files, RELATIONS/*{userId}*.md, and configuring sandboxing via openclaw.json. These requirements align with a multi-user workspace manager and there are no unexplained environment variables, binaries, or external services requested.
Instruction Scope
Instructions are specific and limited to session identification, reading registry/profile/relationship files from the workspace, and configuring sandbox and routing in openclaw.json. They do not instruct the agent to read unrelated system files, exfiltrate data, or call external endpoints. Minor caution: the skill relies on extracting userId from session keys (session_status) — ensure session keys are trusted and well-formed to avoid accidental impersonation or misattribution.
Install Mechanism
This is instruction-only with no install spec and no code files, so nothing is written to disk by the skill installer. That is the lowest-risk install model and is coherent with the described functionality.
Credentials
No environment variables, credentials, or config paths are requested. The only configuration referenced is openclaw.json and workspace subdirectories under the user's home (e.g., ~/.openclaw/workspace). These are proportional to configuring per-user sandboxes and access controls.
Persistence & Privilege
The skill does not request always:true or elevated platform privileges. It does describe configuring agents/sandboxes via openclaw.json (normal for a workspace manager) but does not instruct modifying other skills or system-wide settings outside its own workspace scope.
Assessment
This skill is internally consistent, but review a few operational details before enabling it:
- Verify openclaw.json bindings and sandbox settings carefully. Docker bind mounts like "~/.openclaw/workspace/guests/bob:/workspace:rw" give the container access to the host path; ensure binds only expose the intended directories.
- Limit the number of administrator-role users and audit who can be assigned the administrator Role in USER.md.
- Review FRIENDS/RELATIONS files for any sensitive data before putting real user data into the workspace; those files are read by the assistant and by configured agents.
- Be cautious with allowed tools in agent config (e.g., allowing "exec" or "process" permits running processes inside the sandbox). If you want stricter isolation, deny exec/process or enforce tighter sandboxing.
- Ensure session keys (used to derive userId) come from a trusted source and cannot be spoofed; malformed or attacker-controlled session keys could cause misattribution.
- Test the setup with non-sensitive accounts first and monitor logs to confirm access boundaries behave as expected.
If you want a deeper review, provide your openclaw.json and an example of your USER.md / FRIENDS/RELATIONS files so I can point out any policy or content concerns specific to your configuration.Like a lobster shell, security has layers — review code before you run it.
latestvk970zhmrzaw7a0kha4ba3b3g398133j5
License
MIT-0
Free to use, modify, and redistribute. No attribution required.