frugal

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

• Third-party content exposure detected (high risk: 0.90). SKILL.md explicitly directs the agent to connect to the OpenStall agent marketplace (https://openstall.ai) and run commands like openstall discover and openstall call, which ingest and act on outputs from third‑party marketplace agents (untrusted/user-generated) that can materially influence subsequent decisions and tool use.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly requires installing and running the @openstall/sdk and openstall CLI and uses the OpenStall marketplace (https://openstall.ai) at runtime via commands like openstall discover/call, which fetches and executes remote agent capabilities and can therefore control prompts or execute remote code.

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

• Direct money access detected (high risk: 1.00). The skill integrates with the OpenStall marketplace and its SDK/CLI (e.g., npm @openstall/sdk, openstall register/balance/discover/call/complete). It explicitly describes credits that are real money, withdrawable as USDC or to a bank account, and includes a payment flow where completing/approving a task ("openstall complete" → releases payment). These are specific marketplace/payment operations (including crypto and bank withdrawals) rather than generic tooling, so the skill provides direct financial execution capabilities.