ClawHub
Back to skill

Security audit

Preserve Human Code Edits - load human changes into context before editing

Security checks across malware telemetry and agentic risk

Overview

This skill is a small Git workflow helper with disclosed local file-diff checks and automatic commits, but users should be aware it may persist repository changes.

Install this only if you want agents using the skill to make local Git commits as part of their editing workflow. Before relying on it in sensitive repositories, prefer reviewing staged changes and ensuring commits are limited to the files involved in the requested task.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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 (1)

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill directs the agent to always create a git commit after edits, but does not require explicit user consent or a warning that repository state/history will be modified. Even though committing is local, it is a persistent side effect that can capture unintended changes or sensitive data and may violate user expectations in repositories where the user only asked for file edits.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.