ClawHub
Back to skill

Security audit

rafter-security

Security checks across malware telemetry and agentic risk

Overview

The skill appears to be a legitimate security helper, but its command-validation instructions could cause real commands to run when users may expect only a safety check.

Install only if you are comfortable with Rafter adding agent command-validation hooks and potentially sending code to Rafter for remote scans when using API-backed features. Until the documentation is clarified, use non-executing dry-run command checks for review/classification and do not route destructive or privileged commands through `/rafter-bash` unless you intend them to run and have separately approved them.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
88% confidence
Finding
The documentation advertises broad protections such as output filtering and audit logging, but the concretely described commands in this skill mostly expose scanning, auditing, and validation interfaces without showing how those protections are actually enforced. This can mislead users into over-trusting the skill's coverage and safety guarantees, creating a security gap where risky operations may proceed under false assumptions.

Intent-Code Divergence

High
Confidence
93% confidence
Finding
The metadata describes `rafter agent exec --dry-run` as the command-risk classification interface, but the `/rafter-bash` section instructs users to run `rafter agent exec <command>` directly. If `exec` actually performs execution rather than pure classification, this documentation could cause users or agents to run dangerous commands while believing they are only validating them.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.