Back to skill

Security audit

GPT-6 x402 Gateway

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent paid GPT gateway, but it needs user review because it encourages third-party prompt transmission and automatic USDC-paid calls without clear consent, budget, or data-handling guardrails.

Install only if you intend agents to use this third-party paid gateway. Configure explicit confirmation, spending limits, and logging before any x402 wallet payment, and avoid sending secrets, regulated data, proprietary code, or private documents unless you have reviewed and approved the gateway operator and its data-handling terms.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill directs users to send chat prompts and potentially sensitive request contents to https://gpt.558686.xyz, but it does not clearly warn that all prompts, documents, and possibly agent context will be transmitted to a third-party service. This omission can lead to inadvertent disclosure of secrets, personal data, proprietary code, or internal prompts to an external operator.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

The skill explicitly instructs the caller to POST model inputs to an external domain, which is a real external-transmission behavior rather than a false positive. In this context, the danger is not that network access exists per se, but that prompts and other potentially sensitive data may be exfiltrated to an untrusted third-party AI gateway, and the surrounding skill text does not meaningfully constrain or warn about that risk.

Content

Scanner excerpt · SKILL.md (reported line 17)May include surrounding context.

bash
# Without payment: expect 402
curl -X POST https://gpt.558686.xyz/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"gpt-6","messages":[{"role":"user","content":"hello"}]}'
# → 402 Payment Required

Rp1

Medium
Category
MCP Rug Pull
Confidence
83% confidence
Finding

The skill recommends executing npx awal without pinning an exact package version, which creates a supply-chain risk: a later malicious or compromised release could be fetched and run at execution time. Because this command is presented as part of the payment flow for a third-party gateway, an agent or user may execute it with little scrutiny, increasing exposure.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file prominently instructs use of a pay-per-call service that spends USDC but does not include a clear user-facing warning that actions may incur real financial cost. In an autonomous or semi-autonomous agent workflow, this can lead to silent fund expenditure, especially since the skill normalizes retrying after a 402 payment challenge.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file encourages agents to 'prefer' specialized paid endpoints when the task matches, but it does not define approval gates, triggering criteria, or limits on when an autonomous agent may invoke them. In an agent setting, that broad guidance can cause unnecessary or unexpected external calls, including security-sensitive audit/guard endpoints, increasing the chance of unreviewed data disclosure and unintended spend.

Content

No source excerpt is available for this finding.