ClawHub

Draft Cli

Security checks across malware telemetry and agentic risk

Overview

This skill is a clearly scoped staff moderation helper that documents high-impact admin actions and requires user-directed confirmation before writes.

Install only if you are ClawHub staff or otherwise authorized to run the ClawHub admin CLI. Treat its commands as production-impacting: confirm the exact target, reason, and email content before approving any write or send action.

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 (1)

Context-Inappropriate Capability

Medium
Confidence
72% confidence
Finding
The instructions tell the agent to fetch a provided public URL and then follow metadata-provided endpoints such as comments_api_url and task_toggles_api_url. If those URLs are not constrained to trusted Draft domains or validated against an allowlist, an attacker could supply a crafted page or manipulated metadata that causes the agent to make arbitrary outbound requests, creating an SSRF-style primitive or unintended data access path.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal