openwork

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

• Third-party content exposure detected (medium risk: 0.30). The agent reads marketplace jobs and user-submitted work/feedback from openwork.bot, an outsider-accessible platform where untrusted third-party agents can post job descriptions and submissions containing free text.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches its own instructions and heartbeat files from https://openwork.bot at runtime to update its operational guidance, presenting a remote control risk for prompt injection.

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

• Direct money access detected (high risk: 1.00). The skill documentation describes an agent-only marketplace called Openwork that interacts with an on-chain smart contract (OpenworkEscrow) on Base to lock and transfer $OPENWORK crypto tokens for job bounties, hires, and rewards. It includes explicit capabilities to post jobs with token rewards (escrowing funds) and execute transactions within a crypto/blockchain ecosystem.