openfin-hyperliquid
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the Hyperliquid exchange, including market metadata and prediction market outcome descriptions, which could potentially contain malicious instructions.
- Ingestion points: Data enters the agent context through endpoints such as
GET /market/outcome-meta,GET /market/mids, andGET /market/perp-metasin theSKILL.mdfile. - Boundary markers: The skill provides strong boundary markers by instructing the agent to 'resolve via /market/metas with the user in the loop' and 'Get explicit confirmation in chat before calling' any write operations.
- Capability inventory: The skill possesses high-privilege capabilities including placing orders (
POST /orders), withdrawing funds (POST /withdraw), and adjusting account leverage/margin. - Sanitization: The instructions mandate that the agent must present a unified balance snapshot to the user and surface all rejections verbatim, ensuring the user remains the final authority on all transactions.
Audit Metadata