okx-mcp-services
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from news and Twitter, which is a potential vector for indirect prompt injection attacks where malicious instructions are hidden in data.
- Ingestion points: Data is retrieved via several operations including
news-search,twitter-search,twitter-user-tweets, andtwitter-article-by-idas documented inreferences/newsliquid/http-api.md. - Boundary markers: The instructions in
SKILL.mdandreferences/newsliquid/newsliquid-ws-protocol.mdexplicitly direct the agent to treat external content as data and not as instructions or authorization for payments, which helps mitigate accidental execution. - Capability inventory: The skill utilizes the
onchainosCLI to perform sensitive tasks such as wallet login, billing management, payment authorization, and monitor configuration (references/newsliquid/client-cli-reference.md). - Sanitization: The skill employs clear prompt-level constraints and requires explicit user consent for any state-changing or financial operations, reducing the risk of the agent obeying instructions embedded within external data.
Audit Metadata