The Agent Skills Directory

[SAFE]: The skill explicitly prohibits the handling, storage, or transmission of private keys or seed phrases, mandating the use of trusted signer boundaries or approved API wallets.
[SAFE]: Implements a mandatory human-in-the-loop authorization gate, requiring users to explicitly approve actions by ticket ID after a final review turn.
[SAFE]: Includes a comprehensive suite of pre-send validation checks (pre-send-checks.md) that verify ticket integrity, network settings, account addresses, and market conditions against live exchange metadata immediately before execution.
[SAFE]: The operational scope is strictly limited to trading actions (orders, cancels, leverage) and explicitly excludes high-risk administrative functions such as withdrawals, transfers, or account modifications.
[SAFE]: Provides a detailed incident response framework and state machine to handle edge cases like timeouts or unexpected position changes without improvising potentially dangerous write operations.

hyperliquid-operate