The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell command templates for auditing tasks, including grep, forge test, and environment checks for anchor and solana CLIs. If the agent executes these commands using arguments (such as file paths or contract names) provided by an untrusted user without proper sanitization, it could be vulnerable to command injection.
[DATA_EXFILTRATION]: The instructions reference using a MAINNET_RPC environment variable for blockchain forking. While standard for DeFi development, agents must ensure environment variables are not inadvertently exposed or pointed to malicious endpoints controlled by an attacker during the auditing process.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of Solidity and Rust smart contract source code.
Ingestion points: The agent reads external source code files (e.g., from a contracts/ directory).
Boundary markers: No specific delimiters or instructions are provided to distinguish between actual code and potential malicious instructions hidden in code comments or metadata.
Capability inventory: The skill utilizes shell commands (grep), test runners (forge), and blockchain CLI tools that interact with the local file system and network.
Sanitization: There are no explicit instructions for the agent to sanitize or validate the content of the contracts before processing them with the provided grep patterns or test templates.

web3-audit