Executing Plans
Security checks across malware telemetry and agentic risk
Overview
This is an instruction-only planning workflow that matches its stated purpose and does not ask for credentials, persistence, hidden execution, or unrelated access.
Safe to install as a lightweight workflow skill. Use it only with implementation plans you trust, read each checkpoint before approving more work, and review the referenced finishing-a-development-branch skill separately because final behavior depends on it.
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.