Agent Harness

• Outcome: Review
• Latest audit: Jun 10, 2026 at 5:27 PM UTC
• Version:
  2.0.1

Overview

This is a legitimate code-review helper, but its default helper can run a nested reviewer with full local access and may send code diffs to external review tools.

Install only if you are comfortable with an autoreview helper launching nested reviewer tools over your repository. For sensitive codebases, use --no-yolo or AUTOREVIEW_YOLO=0, set --fallback-reviewer none if you do not want diffs sent to other reviewer CLIs, and review the helper command before allowing an agent to run it automatically.

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: 93% confidence
• Finding: The skill description includes broad, everyday trigger terms such as 'build', 'plan', 'spec', 'review', 'ship', and 'debug', which are common in many unrelated conversations. This can cause unintended activation of a powerful orchestration skill, increasing the chance that users are routed into an over-privileged or inappropriate workflow without explicit intent.

VirusTotal

64/64 vendors flagged this skill as clean.