ClawHub
Back to skill

Security audit

Error Guard — Control‑Plane Safety

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed emergency control skill, but it can terminate active sessions broadly and includes an under-documented sub-agent spawning helper.

Install only if you need an advanced control-plane recovery tool and trust the callers who can invoke it. Review or restrict /flush and /recover before autonomous use, and consider disabling or scoping the sub-agent spawning helper unless you explicitly need delegated long-running work.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
This helper exposes a general-purpose sub-agent spawning primitive, which exceeds the narrowly described recovery/control-plane role in the metadata. In a safety-oriented skill, the ability to launch isolated long-running work materially expands execution capability and can be abused to run arbitrary delegated tasks, increasing the attack surface and weakening operator assumptions about what the skill can do.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The flush operation unconditionally enumerates and kills all active exec sessions as a side effect of an emergency control command, with no confirmation step, scoping, or authorization guard visible in this module. In a control-plane skill, that makes accidental or unauthorized invocation capable of disrupting unrelated work and causing denial of service beyond the intended task registry cleanup.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.