AuctionClaw

• Outcome: Review
• Latest audit: May 29, 2026 at 9:25 PM UTC
• Version:
  2.1.0

Overview

This skill appears to be a real 638Labs routing integration, but it is too broad about sending ordinary tasks and prompt content to an external service without clear consent and credential-safety guidance.

Install only if you intentionally want selected tasks routed through 638Labs. Avoid sending secrets, customer data, private code, regulated data, or confidential files unless you are comfortable with external processing, and store or rotate the API key carefully.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (4)

Vague Triggers

Medium

• Confidence: 94% confidence
• Finding: The skill makes the auction tool the default for essentially any request that does not explicitly name an agent or ask for comparison, which can cause broad unintended routing of user tasks to an external third-party service. This increases the chance of accidental data disclosure, unnecessary external processing, and surprising tool invocation for routine prompts that the user may not expect to leave the local/system context.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The category inference rules use very common phrases such as 'explain', 'analyze', 'help me think through', and 'chat', which are broad enough to match many ordinary conversations and trigger external auction routing. In this skill's context, that means benign user dialogue may be silently transformed into third-party processing, increasing privacy and consent risks.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The setup flow instructs the agent to ask for an API key and save it to a local env file, but provides no warning about sensitivity, scope, file permissions, rotation, or safer secret-handling practices. This can lead to credential exposure through insecure storage, accidental logging, or users pasting secrets into chat without understanding the risk.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill description markets broad task routing but does not clearly disclose that prompts and possibly attached task data will be sent to an external HTTP MCP endpoint for processing. Because this skill covers scraping, code, chat, translation, and other potentially sensitive workloads, lack of upfront disclosure materially increases the risk of users unknowingly sending confidential data to a third party.

VirusTotal

66/66 vendors flagged this skill as clean.