The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external blockchain-related sources, creating a potential surface for indirect prompt injection attacks.\n
Ingestion points: The scripts/evm_client.py script fetches data from blockchain RPCs (transaction input, token names), 4byte.directory (function signatures), and ensideas.com (ENS names, social media handles, and profile descriptions).\n
Boundary markers: The skill returns results in a structured JSON format, but it lacks specific instructions or markers to prevent the agent from following instructions potentially embedded within the retrieved string data.\n
Capability inventory: The skill utilizes the terminal toolset and the Python urllib module for network communication. It lacks high-risk capabilities such as file-system modification or privilege escalation.\n
Sanitization: No sanitization, escaping, or filtering is performed on the text retrieved from external APIs and blockchain state before it is returned to the agent's context.

evm