ClawHub
Back to skill

Security audit

Research Loop

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed research workflow skill that can write local research artifacts and publish them to a chosen GitHub repository, but that behavior matches its stated purpose.

Install this only when you are comfortable letting an agent create files for a research workflow and potentially publish sanitized artifacts to the target GitHub repository. Use the --no-push or local-only path for dry runs, review the .autonomous-research-workflow and source_changes outputs before publishing, and avoid running it against private or sensitive repositories unless the GitHub CLI credentials are scoped appropriately.

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

Low
Confidence
93% confidence
Finding
The document explicitly states that each run writes durable local state and publishes artifacts into the target repository, but it does not present any user-facing warning, consent requirement, or guardrail around those filesystem modifications. In an agent skill, undocumented writes to a local working tree or repo can surprise users, alter tracked files, and create unintended commits or leakage of run artifacts into project history.

VirusTotal

58/58 vendors flagged this skill as clean.

View on VirusTotal