adversarial-coach

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only code review skill whose behavior matches its stated purpose and shows no hidden execution, persistence, exfiltration, or unrelated authority.

Use it with a clear requirements file or issue reference so the review scope is bounded. For untrusted repositories, treat any build, test, or run command as normal code execution risk and approve those steps deliberately.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (1)

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The skill advertises `/coach` with optional arguments and explicitly permits inferring requirements from surrounding context when no file is provided. That broad activation and implicit context harvesting can cause the agent to review or act on unintended materials, increasing the chance of prompt-scope confusion, accidental disclosure of nearby sensitive context, or misuse in conversations where the user did not clearly delimit the review target.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal