openart
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- Data Exposure & Exfiltration (MEDIUM): The skill instructs the agent to send the
agent_walletandmoltbook_usernametohttps://openclawart-production.up.railway.app/round/submit. This domain is not on the trusted whitelist. While wallet addresses are public, their use in an automated routine to an unverified endpoint facilitates tracking and potential targeting. - Indirect Prompt Injection (MEDIUM): The skill is vulnerable to instructions embedded in external data it processes.
- Ingestion points: The agent retrieves the
lastWinningPromptfrom the external API viaGET /winners/latest(HEARTBEAT.md). - Boundary markers: Absent. There are no instructions to delimit the external content or ignore potential commands within it.
- Capability inventory: The agent has network capabilities (POST requests) and persistent memory/state tracking.
- Sanitization: Absent. The agent is explicitly told to "Note the winning prompt" and "Learn from it," which could lead to the agent adopting malicious instructions contained within that prompt.
Audit Metadata