The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell-based tools including curl, jq, and cast for blockchain interactions and API queries. It implements robust security controls by mandating strict regex validation for all user-supplied variables (such as token addresses and amounts) and explicitly instructing the agent to reject any input containing shell metacharacters, effectively preventing command injection.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risks of processing untrusted data from web searches and third-party APIs like DexScreener. It mitigates potential indirect prompt injection attacks by requiring the agent to label discovered tokens as untrusted, provide security warnings, and obtain explicit user confirmation via the AskUserQuestion tool before taking action.
[EXTERNAL_DOWNLOADS]: The skill fetches token metadata, price data, and contract bytecode from public APIs and blockchain RPC nodes. These operations are limited to data retrieval from well-known services and do not involve the execution of remote scripts or binary downloads.
[SAFE]: The skill uses the system's default browser opener (xdg-open or open) to facilitate the swap on the official Uniswap application. This behavior is restricted to the intended purpose of the skill and is protected by the aforementioned input validation layers.

swap-planner