The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill facilitates the ingestion of external data which could potentially contain malicious instructions intended to manipulate the agent's behavior.
Ingestion points: Market data (tickers, trades), account data (bills, fills), and especially error messages (msg field) defined in references/okx-error-map.md and references/okx-sdk-usage.md.
Boundary markers: There are no explicit instructions or delimiters defined to treat API-returned strings as untrusted data or to ignore instructions embedded within them.
Capability inventory: The skill provides methods for full trade execution, including submitOrder, cancelOrder, and setLeverage as seen in references/okx-sdk-usage.md.
Sanitization: No sanitization or validation logic is prescribed for strings returned by the OKX API before they are presented to the agent's reasoning context.
[Data Exposure] (LOW): The skill instructions in SKILL.md explicitly require the use of sensitive credentials (API key, secret, and passphrase). While no hardcoded secrets are present, the handling of these high-value credentials by the agent creates an inherent exposure risk.

okx-provider-broker