Swarm
Security checks across malware telemetry and agentic risk
Overview
The skill looks like a legitimate parallel research helper, but it gives agents broad shell and local-history/code-search access without clear limits.
Review the plan before allowing the skill to run. Consider telling the agent not to use Bash unless you explicitly approve a command, and specify whether it may search past sessions, local knowledge bases, or project code. Check docs/research.md before sharing it because it could include private local or project information if those sources are enabled.
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
65/65 vendors flagged this skill as clean.