ClawHub
Back to skill

Security audit

skill-studio

Security checks across malware telemetry and agentic risk

Overview

This is a coherent skill-building toolkit, but it needs Review because its installer can overwrite agent skill folders and its promised validation gate can fail open if the validator is missing or crashes.

Install only if you are comfortable with a skill that can create files, package skills, and copy them into agent skill directories. Before using --target or --target all, review the destination paths and consider fixing package_skill.py so packaging fails if validate.py is missing or crashes.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill explicitly instructs use of local scripts that can read files, write files, package artifacts, and invoke validation/packaging tooling, which implies file_read, file_write, and likely shell/process execution capabilities. Because these capabilities are present but not declared, users and policy enforcement layers may not realize the skill can modify the workspace or execute commands, creating a transparency and least-privilege problem rather than clear malicious behavior.

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
The documentation promises that validation failures block packaging, but the implementation treats a missing validator or validator execution failure as success. An attacker or careless operator could remove, rename, or break validate.py to bypass safety checks and package/install an unvalidated skill.

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
The code labels validation as a hard gate, but earlier logic allows packaging to continue when the validator is absent or crashes. In a skill-packaging tool, this context increases risk because validation is a primary control preventing unsafe or malformed skills from being distributed and installed across multiple agent hosts.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.