ALab Experiment Worker

• Outcome: Pass
• Latest audit: Jun 7, 2026 at 2:15 PM UTC
• Version:
  0.1.8

Overview

This skill is a scoped ALab experiment-worktree helper that can edit candidates, run evaluations, and submit results, with clear boundaries against admin/root access and secret handling.

Install this only where you want an agent to operate inside an ALab experiment worktree and potentially edit candidate source, run evaluations, create visible notes, and submit final results. Keep worktree tokens scoped, do not provide admin/root credentials, and consider disabling implicit invocation if you want every use of this high-impact workflow to be explicitly requested.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The skill enables allow_implicit_invocation without defining a narrowly scoped trigger, which increases the chance the agent will invoke this skill in loosely related contexts. Because this skill can inspect evidence, edit source, run evaluations, and submit experiment results, unintended invocation could cause unauthorized or premature actions within an ALab worktree token context.

VirusTotal

64/64 vendors flagged this skill as clean.