ClawHub
Back to skill

Security audit

Team Sessions

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a multi-agent OpenClaw setup skill, but it asks for broad agent delegation and administrative config changes without enough scoping or rollback guidance.

Review before installing. Use only if you intend to change OpenClaw agent configuration, replace wildcard sub-agent permissions with explicit agent IDs where possible, back up openclaw.json first, and be prepared to roll back or restart the gateway if routing or delegation behaves unexpectedly.

SkillSpector

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
95% confidence
Finding
The skill explicitly recommends `allowAgents: ["*"]`, which broadly authorizes spawning any configured subagent rather than only the narrowly needed team members. In a multi-agent environment, this expands the control surface and can enable unintended delegation to more privileged or inappropriate agents if they are present now or added later.

Intent-Code Divergence

Medium
Confidence
83% confidence
Finding
The document states the main agent should not modify files or execute commands, but earlier sections instruct exactly those operations for setup and operation, including editing config and restarting the gateway. This contradiction is dangerous because it obscures the real authority the skill expects, making operators more likely to approve risky actions under a false sense of safety.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill instructs users to modify `openclaw.json` and restart the gateway, both of which are system-impacting administrative actions, without strong safety framing or rollback guidance. Changes to agent permissions and a service restart can alter trust boundaries, break existing workflows, or unintentionally enable broader agent capabilities.

Static analysis

No suspicious patterns detected.