ClawHub
Back to skill

Security audit

Coding Team Setup

Security checks across malware telemetry and agentic risk

Overview

This appears to be a real OpenClaw team setup wizard, but it can broadly rewrite agent configuration and its mandatory telemetry guidance is under-scoped for sensitive project context.

Install only if you are comfortable with a local wizard rewriting OpenClaw agent configuration. Back up openclaw.json first, prefer named teams with --team, review the diff after running, and do not enable raw context telemetry or weekly optimization unless you add redaction, retention limits, and rollback controls.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The documented 'standard post-setup workflow' expands a team setup skill into ongoing cross-team telemetry collection, automated analysis, and configuration rewriting. That creates a real security and privacy risk because it normalizes persistent monitoring and privileged automation beyond the minimum needed to create teams, increasing the blast radius if logs or the optimizer are misused or compromised.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The banner/docstring presents the script as a setup wizard, but the implementation backs up and then rewrites ~/.openclaw/openclaw.json, removing non-default agents for the default team and removing matching-prefixed agents for named teams. This mismatch is dangerous because users may reasonably expect additive configuration, while the script performs destructive state changes that can silently delete prior agent entries and alter main-agent allowlists.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill explicitly instructs logging `agent_id + skill_name + timestamp + context`, and 'context' can contain sensitive prompts, code, secrets, business data, or user information. Requiring this telemetry without a clear privacy notice, data minimization, or retention controls creates a genuine confidentiality risk and may violate user expectations or policy.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal