Back to skill

Security audit

crypto-funding-scanner

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a suspicious skill because it permits paid data requests without asking first, which could lead to repeated charges despite a per-task cap.

Before installing, consider whether you are comfortable with paid requests being made without asking each time. If the connected agent has payment access, repeated tasks could incur cumulative charges; the scanner finding recommends having an explicit user-approved payment policy and spending limits.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The skill authorizes spending wallet funds without asking, but a skill-defined allowance does not establish the user's consent to paid requests. The legitimate paid-data workflow, $0.25 per-task cap, and recipient/network checks substantially limit the risk, yet repeated tasks could still incur unauthorized cumulative charges if the agent has payment access.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

md
## Paying safely (read before any paid call)
- A paid call first answers HTTP 402 with a quote. Pay only if the quote's `payTo` is **0xDeA94ef76afa86bf1638B645F76108dF65890c0b**, the network is Base (`eip155:8453`) and the asset is USDC. Otherwise stop and tell the user.
- Spend at most **$0.25 per task** without asking. Above that, tell the user the total and wait for approval.
- If you have no x402 wallet, use the free sample and say that live data needs an x402-capable client.
- HTTP 503 or an error means no fresh data: say so. **Never invent or estimate numbers.** Failed calls are never charged.
- This is observational market data, not financial advice. Say so when relaying it.