Agent Ops Hardening
Security checks across malware telemetry and agentic risk
Overview
This skill is a local OpenClaw hardening checklist with an optional audit script that checks workspace files and prints recommendations without making changes or sending data.
Safe to install as a hardening aid. Run the audit script only in a workspace you intend to inspect, and treat the external-service pre-flight advice as manual guidance, preferably using read-only or low-risk test calls.
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
VirusTotal
66/66 vendors flagged this skill as clean.