Security audit
Paper Reading Method Commitment
Security checks across malware telemetry and agentic risk
Overview
This is a structured paper-method workflow that only directs local markdown artifact creation and review, with no hidden code, credential use, network access, or unsafe automation found.
This skill appears safe to install for its intended research-writing workflow. Expect it to create or update local method-commitment markdown files under a workspace directory and to ask for user confirmation before treating a method as committed. Be aware that the package references template files that were not included, so the agent may need to reconstruct those template structures from the main instructions.
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
62/62 vendors flagged this skill as clean.