ClawHub
Back to skill

Security audit

Agent Assistant Free

Security checks across malware telemetry and agentic risk

Overview

This skill appears to be a persona-design Markdown guide, but it asks for broad file and command permissions that are not clearly needed or scoped.

Review before installing. The content is not malicious on its face, but the requested command and file-write permissions are broader than a Markdown persona-design helper should need. Prefer installing only if the publisher narrows the trigger conditions and removes or clearly scopes write/exec permissions with explicit user confirmation.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill declares read/write/exec/glob/grep capabilities even though its stated purpose is persona definition and structured prompting, which should not require file modification or shell execution. Over-privileged tool access increases the blast radius if the skill is invoked unexpectedly or if downstream prompts cause the agent to perform filesystem or command actions unrelated to the documented function.

Intent-Code Divergence

Medium
Confidence
91% confidence
Finding
The documentation says the skill has no extra runtime dependency and is 'pure Markdown,' but elsewhere it declares exec and write capabilities and even labels itself MD+EXEC. This mismatch can mislead users and platforms into underestimating operational risk, making unsafe execution or file changes more likely because the dangerous capabilities are obscured by contradictory documentation.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The trigger condition is broad enough to match generic AI-model, agent-orchestration, and LLM-application scenarios rather than a narrowly scoped persona-definition use case. This can cause the skill to activate in unrelated contexts, where its overbroad tooling and identity-shaping behavior may interfere with other workflows or influence agent behavior unexpectedly.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill advertises write and exec capabilities without clearly warning users that it may modify files or run commands. Hidden or poorly disclosed high-risk actions undermine informed consent and can lead to unintended local changes or command execution if the skill is selected based only on its benign-seeming description as a persona-design tool.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.