ClawHub
Back to skill

Security audit

agentskills-io

Security checks across malware telemetry and agentic risk

Overview

This looks like a legitimate Agent Skills authoring and validation helper, but users should treat its validator commands as trusted code execution because they fetch and run code from GitHub.

Install only if you trust the agentskills upstream validator and are comfortable running GitHub-hosted code locally. Prefer pinning the validator to a commit or reviewed release for sensitive environments, and use the templates with added safety sections for permissions, external side effects, approvals, verification, and rollback.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Vague Triggers

Medium
Confidence
87% confidence
Finding
The template’s description guidance encourages broad discovery wording without requiring precise scope boundaries, which can lead authors to create skills that are invoked in inappropriate contexts. In an agent ecosystem, ambiguous invocation metadata increases the chance that an agent selects a powerful or risky skill when a narrower, safer tool should have been used.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The template includes examples involving deployment, cloud credentials, and operational commands, but it does not instruct authors to disclose safety-impacting behavior such as external side effects, infrastructure changes, permission requirements, or rollback steps. This omission can normalize incomplete skill documentation, making it easier for downstream agents or users to execute impactful actions without adequate warning or verification controls.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The script executes a tool fetched directly from a remote GitHub repository via `uvx --from git+https://...` during validation, which introduces a supply-chain risk and undisclosed network/code execution. If the upstream repository, a dependency, or the transport context is compromised, running this validator can execute attacker-controlled code on the user's machine without any explicit warning or pinning to a trusted revision.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.