agentic-wallet

Warn

Audited by Snyk on Nov 21, 2026

Risk Level: MEDIUM
Full Analysis

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (low risk: 0.10). The required path fetches first-party Coinbase docs (coinbase.com/skill.md and the Wallet MCP guide) and then calls Wallet MCP tools, which may return third-party-influenced on-chain data that this skill doesn't describe, so I'm scoring it low under the recall-first override, though the legacy outsider-text references are explicitly excluded.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

  • Direct money access detected (medium risk: 0.30). Direct financial operation, delegated to Wallet MCP. The skill's description (lines 4-6) routes sending, token transfers, swaps, funding, and x402 paid-service requests to Wallet MCP, and lines 27-31 describe withdrawing and transferring funds. Because the skill offers payment authority and transaction routing, this is in scope and direct, not indirect. Its own instructions (lines 17-25) tell the agent to connect to the remote Wallet MCP server and use its live tools for wallet operations. Access control: the skill names Coinbase Wallet OAuth and required approvals for this route (lines 20-22), with the remote server at wallet-mcp.coinbase.com. That is a service-side authentication mechanism for the route, not a prompt-level warning. The OAuth details live in the external guide, which was not supplied, so the strength of the approval step is not verified. I rate this medium because the named service authentication gates the route, but the evidence is partly external. The skill itself does not perform transfers and is a deprecated redirect.

MEDIUM W013: Attempt to modify system services in skill instructions.

  • Attempt to modify system services in skill instructions detected (low risk: 0.10). Gate 1 is passed by the Wallet MCP setup and the local wallet session commands (login, verify, logout), which change the supporting agent integration and session state. Gate 2 fails because no explicit unrecoverable loss of operational state is described: logout is recoverable by signing in again, and the wallet migration is directed to the user's own withdrawal flow. The result is therefore low, not medium or high.

Issues (3)

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W009
MEDIUM

Direct money access capability detected (payment gateways, crypto, banking).

W013
MEDIUM

Attempt to modify system services in skill instructions.

Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 21, 2026, 08:51 PM
Issues
3