Security audit
Harness Writing
Security checks across malware telemetry and agentic risk
Overview
This is a documentation-only fuzzing-harness guide with no hidden access, install hooks, or persistent privileges.
This skill appears safe to install as a reference for writing fuzzing harnesses. Users should still review generated harness code before compiling or running fuzzers against local projects, because fuzzing intentionally executes project code with malformed inputs.
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
63/63 vendors flagged this skill as clean.