ClawHub
Back to skill

Security audit

neon-functions

Security checks across malware telemetry and agentic risk

Overview

This skill is coherent Neon Functions documentation, but needs review because some copyable examples expose public database-backed endpoints or third-party telemetry before clearly scoping access and sensitive data handling.

Review this skill before installing if you will let an agent deploy code. Treat every Neon Function URL as public: require authentication before database, MCP, WebSocket, SSE, or agent work; scope database actions to the authenticated user or tenant; sanitize telemetry before sending it to Sentry or Mastra; and avoid using the broad triggers unless you specifically intend to build on Neon Functions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The primary example publishes a public `/mcp` endpoint with state-changing tools (`create_contact`, `delete_contact`) and no authentication or authorization check in the handler. Although the document later warns that MCP servers must authenticate callers, many users copy the first complete example as-is, which can lead to unauthorized database reads/writes and abuse of all exposed tools.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The trigger list includes broad, common phrases such as 'serverless function', 'deploy an API', and 'webhook handler', which can cause the skill to activate in situations where a user did not specifically intend to use Neon Functions. Because this skill can steer users toward deploying public HTTP endpoints and database-adjacent infrastructure, unintended invocation can lead to mis-scoped assistance, accidental exposure, or insecure deployment guidance being applied in the wrong context.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The documentation explicitly recommends exporting agent traces to Mastra Studio but does not warn that traces may include prompts, model inputs/outputs, tool arguments, or other sensitive application data. In an agent-hosting skill, this omission can lead users to unknowingly transmit regulated or secret data to a third-party observability service, creating confidentiality and compliance risk.

Missing User Warnings

Medium
Confidence
84% confidence
Finding
The example recommends attaching `url` and `source` as Sentry event context without warning that these fields may contain sensitive data and will be transmitted to a third-party monitoring service. In an agent/webhook/API context, URLs can embed query secrets, tokens, PII, or internal hostnames, so copying them verbatim into telemetry can create unintended data exposure outside the Neon environment.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.