Back to skill

Security audit

magneta-contract-risk

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because it reports automatic paid requests without a clear user-approval check, including a payment check that may allow an unapproved offer through.

Consider the payment approval gap and the possibility that a mixed payment request may pass the callback. The scan is partial: one static analysis was degraded and no components were fully inspected, so the evidence is incomplete. The verdict is suspicious, not a claim that the author acted maliciously.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The risk-report request and Base payment mechanism match the stated service. However, the payment-enabled fetch can sign a payment authorization immediately upon receiving a qualifying payment challenge. A per-payment spending limit restricts the amount but does not establish user consent. Although the comment refers to an approved price, the supplied code neither obtains nor verifies approval before making the payable request.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

The risk-report request and service endpoint match the declared purpose. However, the description promises payment only with explicit user approval, while the code treats a hardcoded amount as approved and enables automatic payment. The comment stating that the user approved the price does not establish or enforce consent. The payment checks restrict currency, network, and amount but do not verify approval for this call.

Content

No source excerpt is available for this finding.

MCP payment approval does not enforce the documented asset and spending restrictions for mixed payment offers (L114–L117).

High
Category
Excessive Agency
Confidence
93% confidence
Finding

The MCP callback checks only the filtered Base-USDC offers, then returns approval for the payment request without removing or rejecting other offers. A mixed request containing one acceptable offer and another same-network offer with an unapproved asset or amount can therefore pass this gate; whether the unchecked offer is signed depends on the SDK's selection behavior. Registering only Base limits network exposure, and the dedicated low-balance wallet reduces likely losses, but neither ensures that the selected offer satisfies the user's asset and price restrictions.

Content

No source excerpt is available for this finding.