ClawHub
Back to skill

Security audit

magneta-contract-risk

Security checks for vulnerabilities and agentic risk

Overview

The paid examples can authorize up to $0.05 despite requiring approval for a $0.02 report, so payment controls need review.

Use the free checks without wallet credentials. Before enabling paid reports, require a $0.02 limit in every payment path, restrict payment to USDC on Base, and approve each report separately. Use a dedicated low-balance wallet and pin every optional package version.

A.I.G

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

SkillSpector

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
99% confidence
Finding

The MCP example enables automatic payment and approves requests using an amount threshold of $0.05, rather than the advertised $0.02 price. The surrounding workflow requires prior user approval, so automatic settlement alone is not a demonstrated consent bypass; however, an increased server quote could pass the callback without approval for that increased price. This overlaps the first finding and does not establish a separate malicious payment mechanism.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The payment examples permit charges up to $0.05 despite the stated $0.02 price and explicit instruction to refuse anything higher, allowing a server-supplied payment request to exceed the approved price. Automatic payment is not independently evidence of missing user consent because the workflow explicitly requires approval before running the paid path, but the examples do not enforce that approval's price boundary. The small cap and dedicated low-balance wallet guidance limit financial exposure.

Content

No source excerpt is available for this finding.

Payment checks permit charges above the documented $0.02 price.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The configured $0.05 per-payment ceiling permits a charge above the documented and user-approved $0.02 report price. A compromised or misconfigured service could exploit this discrepancy without violating the client's configured ceiling, although the demonstrated excess is only $0.03 per payment. This is the same underlying price-enforcement weakness identified in the two description-behavior findings.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.