The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Foundry toolkit by piping a remote script from https://foundry.paradigm.xyz into a bash shell. This is a standard installation method for the well-known Foundry development framework.
[EXTERNAL_DOWNLOADS]: The skill describes downloading security tools like echidna and hevm from official GitHub release pages (such as crytic/echidna and ethereum/hevm) and fetches blockchain data from public RPC endpoints like eth.llamarpc.com. It also references online signature databases like 4byte.directory and etherface.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted smart contract bytecode retrieved from the blockchain or local files. Ingestion points: Contract bytecode is obtained via the cast code command or from local binary/hex files. Boundary markers: The instructions do not define specific delimiters for separating untrusted bytecode from the agent's instructions. Capability inventory: The skill utilizes decompilers and disassemblers (panoramix, cast) to analyze code and executes system commands to process the data. Sanitization: There is no mention of sanitizing or verifying the input bytecode before it is passed to the analysis tools.
[COMMAND_EXECUTION]: The skill employs various command-line utilities for blockchain forensics, including cast, panoramix, pip, and system package managers like apt, dnf, and pacman.

re-blockchain