ClawHub
Back to skill

Security audit

AgentGit

Security checks across malware telemetry and agentic risk

Overview

This skill has a coherent validation-and-merge purpose, but it asks users to install an unpinned external GitHub CLI that can merge repository changes, so it should be reviewed before use.

Review the AgentGit repository and pin a trusted commit or release before installing. Use this only in repositories where automated validation commands and branch merges are acceptable, and confirm task approval is fresh before running merge.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (2)

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The skill documentation instructs users to clone and potentially install/link an external GitHub repository to obtain the `agentgit` CLI, which adds network-based code acquisition and trust in third-party code beyond the stated purpose of validating sub-agent output. In a security-sensitive agent skill, this expands the attack surface materially because the fetched tool could be tampered with, drift over time, or execute unexpected code during use or installation.

Intent-Code Divergence

Medium
Confidence
80% confidence
Finding
The document makes a strong safety guarantee that tasks will never be merged unless validated in the same session, but the surrounding text only establishes an approval-state check. If operators rely on this stronger claim, they may assume freshness and session binding that may not exist, enabling stale approvals, mismatched validation context, or unintended merges based on prior state.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.