Trading Co-Agent Pro

• Outcome: Review
• Latest audit: May 29, 2026 at 9:31 PM UTC
• Version:
  1.0.0

Overview

This skill is openly for AI-assisted live trading, but it gives the agent broad authority to place, retry, and remediate real trades without clear user approval or hard limits.

Install only if you intentionally want an AI agent involved in supervised trading automation. Keep it read-only or paper-trading by default, require explicit confirmation for live orders and retries, use tightly scoped trading-only API keys with withdrawals disabled, set hard asset/account/position/loss limits, and ensure there is a clear stop control before granting real trading access.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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

Findings (5)

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The manifest activates on broad conditions such as general trading collaboration, monitoring signals, and executing real trades, without clear boundaries, prerequisites, or limits. This can cause the skill to be invoked in high-risk contexts where it may take action too readily, especially because the rest of the skill grants autonomous trading authority.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The skill description explicitly authorizes real trading, corrective trades, and continuous optimization, but it does not require an explicit user warning or informed consent before performing financially consequential actions. In this context, that omission is dangerous because trading and remediation actions can irreversibly move assets, realize losses, or create additional exposure.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: These instructions direct the AI to proactively execute trades, retry failed trades, and compensate for perceived script shortcomings, all without a mandatory consent gate or human approval step. In a live trading environment, autonomous retries and self-initiated execution can rapidly amplify losses, create unintended positions, or trigger repeated harmful transactions under faulty assumptions.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The emergency remediation section instructs the AI to perform corrective sell or hedge actions in response to mistaken trades, but it omits explicit warnings that these actions can lock in losses, incur fees, and create further exposure. Because these are irreversible asset-impacting operations taken under urgent conditions, the lack of consent and safety constraints materially increases risk.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: This checklist gives prescriptive trading recovery steps such as immediately selling, reevaluating losses, and re-executing trades, but does not include any warning that these actions are discretionary and can worsen losses. In a trading-oriented skill, users may treat the checklist as operational guidance and follow it automatically, increasing the chance of financial harm from impulsive or context-insensitive decisions.

VirusTotal

51/51 vendors flagged this skill as clean.