ClawHub
Back to skill

Security audit

Agentchat Skill

Security checks across malware telemetry and agentic risk

Overview

This chat skill appears purpose-aligned, but it should be reviewed because it can automatically install software, persist new agent instructions, and remove local configuration files.

Review the setup and troubleshooting commands before installing. Only allow the MCP add, npx install, curl download, MCP removal, or rm -rf cleanup steps if you intended to modify your local agent configuration and trust the GitHub repository and npm package source.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill directs the agent to run shell commands that install an MCP server and fetch a remote SKILL.md into the local filesystem. That goes beyond merely using chat functionality and creates a software supply chain and local-environment modification risk, especially because the fetched content is remote and unauthenticated within the skill text.

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The troubleshooting section includes local administrative commands, including removal of MCP configuration and deletion of the local skill directory. These actions are not required for normal chat use and can damage or disrupt the user's local environment if followed automatically or without clear consent.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The installation flow performs a package install plus a network download piped into a local file write, but the skill does not clearly warn about those side effects or the trust assumptions involved. This creates a meaningful risk of unreviewed remote content being introduced into the user's environment under the guise of routine skill setup.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The troubleshooting instructions recommend `rm -rf .claude/skills/agentchat` without an explicit warning that local files will be permanently deleted. Even if scoped to one directory, destructive deletion commands in a skill can be unsafe when followed by an agent or user without careful review.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal