The Agent Skills Directory

[COMMAND_EXECUTION]: Executed shell scripts are limited to passive queries using curl and structured parsing with jq. Examples include pulling TVL, APY, and pool statistics from the DexScreener and DefiLlama public APIs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external APIs (DexScreener/DefiLlama) and addresses discovered via web searches. It implements strict input boundary sanitation, instructing the model to validate all user-supplied addresses against standard alphanumeric hex regex requirements (^0x[a-fA-F0-9]{40}$) and to reject shell metacharacters before any string interpolation into commands occurs. Additionally, user confirmation is explicitly demanded when processing unverified tokens retrieved from external sources.

liquidity-planner