Huizai Error Guard

• Outcome: Review
• Latest audit: May 28, 2026, 6:17 AM
• Version:
  1.0.0

Overview

This error-recovery skill is mostly coherent, but its recovery command can stop all active agent sessions, not just its own work.

Install only where it is acceptable for a recovery action to interrupt active agent sessions. Before using it in shared or production workflows, require confirmation or authorization for flush/recover, restrict process killing to this skill's own spawned sessions, and define redaction rules for any saved error context or event metadata.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Intent-Code Divergence

Medium

• Confidence: 94% confidence
• Finding: The flush command is documented as an emergency stop for task handling, but it also iterates over and kills all active exec sessions globally. If this command is reachable by less-trusted callers, it becomes a broad denial-of-service primitive that can terminate unrelated work outside this task registry's scope.

Missing User Warnings

Medium

• Confidence: 83% confidence
• Finding: This helper unconditionally sends task event data and arbitrary metadata to a remote session via sessions_send without any visible consent, disclosure, filtering, or minimization. If callers place sensitive values in meta or misuse sessionKey, the code can exfiltrate operational or user-related data to another session/channel with little transparency or control.

VirusTotal

66/66 vendors flagged this skill as clean.