Aiter Ck Gemm Tune

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed GPU kernel tuning workflow, but it should only be used in an approved AITER/ROCm environment because it can run remote commands, edit tuning-related files, and launch long jobs.

Install only if you intend to tune AITER kernels on a trusted AMD GPU node or container. Confirm the exact host/container, AITER path, files to be modified, log/report destinations, and any background tuning job before letting the agent proceed; monitor GPU usage and stop long-running jobs when finished.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (2)

Context-Inappropriate Capability

Medium

Confidence: 87% confidence
Finding: The skill instructs the agent to obtain SSH and Docker access details and then run subsequent steps inside that remote environment. While operationally plausible for GPU tuning, this expands the skill from local performance tuning guidance into remote-environment access and command execution, increasing the blast radius if the agent is misused or the instructions are followed automatically.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The skill directs modification of test scripts to relax argparse restrictions, which is a source-code editing action beyond simple tuning execution. Editing test code can weaken safeguards, normalize broader source mutation by the agent, and create unintended or persistent changes in the environment that outlive the tuning session.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal