ClawHub
Back to skill

Security audit

Critic Agent

Security checks across malware telemetry and agentic risk

Overview

This is a coherent critic/reviewer skill; its main risks are expected privacy and retention considerations when sending outputs to a model or logging critiques.

Before installing, decide whether reviewed tasks and agent outputs may contain secrets, customer data, proprietary code, or regulated information. Configure the model/provider deliberately, avoid sending sensitive content unless that provider is acceptable, and log only minimal or redacted critique data with a clear retention policy.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill explicitly describes sending task content and agent outputs to an LLM, but it does not warn users that this data may be transmitted to an external API provider. Because the reviewed content may contain proprietary code, prompts, or sensitive workflow context, users could unknowingly disclose sensitive information outside their trust boundary.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The document explicitly instructs logging all critiques to a persistent file, and the input/output schemas show that critiques may contain the original task, agent output, requirements, constraints, and success criteria. Persisting this material without any warning, minimization, redaction guidance, retention limits, or access controls can expose sensitive prompts, proprietary data, or personal information to later users, tools, or attackers with filesystem access.

Ssd 3

Low
Confidence
84% confidence
Finding
The best-practice guidance to 'Log all critiques' can lead operators to store full prompts, agent outputs, and feedback verbatim, which may include secrets, personal data, or sensitive model-generated content. In a critic skill, this risk is amplified because the system is designed to ingest arbitrary upstream agent content, making sensitive-data retention plausible even if not intended.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.