The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/collect_transaction_data.py uses subprocess.run to execute the etherscan CLI tool for data retrieval. This implementation is secure as it uses list-based arguments without a shell (shell=False) and validates all user-provided inputs, such as transaction hashes and contract addresses, against strict regular expressions (^0x[0-9a-fA-F]{64}$ and ^0x[0-9a-fA-F]{40}$) before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external transaction data and contract metadata from Etherscan, creating a potential surface for indirect injection. However, the risk is minimized by the skill's architecture: it uses structured JSON data for analysis, implements internal logic for fact derivation (e.g., scripts/summarize_transaction.py), and provides the agent with explicit instructions to distinguish between observed facts and inferred intent, including a required confidence-rating system.

etherscan-transaction-debugger