The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests dynamic content from various cryptocurrency exchange APIs and Coinglass without explicit sanitization. This untrusted data is processed in an environment where the agent possesses trade execution capabilities. Ingestion points: funding_rate(), cg_coins_market_data(), and cg_open_interest() tools in SKILL.md. Boundary markers: No delimiters or ignore instructions are present to separate external data from system instructions. Capability inventory: Access to hl_order() for market execution and hl_account() for viewing sensitive wallet balance information. Sanitization: No input validation or content filtering mechanisms are defined for the data returned by the APIs.
[COMMAND_EXECUTION]: The skill utilizes powerful tools to interact with financial infrastructure and sensitive user data. Tool usage: hl_account() is used to retrieve current balance and margin information for position sizing. Tool usage: hl_order() is used to submit trade orders to the Hyperliquid exchange. Safety mitigation: The instructions include a strict rule stating that no trade can be executed without an explicit "YES" from the user, which reduces the risk of autonomous or accidental execution.

@1826/funding-rate-arb