Security audit

Grilling

Security checks across malware telemetry and agentic risk

Overview

This is a small planning-interview skill whose behavior is disclosed and proportionate, with the main caveat that it may inspect the codebase when relevant to answering planning questions.

Install this if you want Codex to challenge plans more aggressively. Be aware that for code-related plans it may read relevant project files to avoid asking questions it can answer itself; review the source URL and prefer pinned or marketplace installs if you want stronger supply-chain control.

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 (2)

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The activation condition is broadly framed around any request to stress-test a plan, without strong constraints on context or user confirmation. In combination with the instruction to ask relentless follow-up questions and to explore the codebase when possible, this broad trigger increases the chance of unwanted activation and unnecessary access to project context.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The activation condition is broadly framed around any request to stress-test a plan, without strong constraints on context or user confirmation. In combination with the instruction to ask relentless follow-up questions and to explore the codebase when possible, this broad trigger increases the chance of unwanted activation and unnecessary access to project context.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.