The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes instructions and script logic that download a setup script from a remote URL and pipe it directly to a shell interpreter.
Evidence: curl -L https://foundry.paradigm.xyz | bash is recommended for installing the Foundry toolkit.
Found in: SKILL.md and the check_dependencies function in scripts/execute-swap.sh.
[COMMAND_EXECUTION]: The skill executes multiple local shell scripts and external CLI tools to build and broadcast blockchain transactions.
Operations: Runs scripts/execute-swap.sh and scripts/fast-swap.sh.
Tool usage: Invokes curl for network requests, jq for data parsing, and cast for wallet management and transaction execution.
[CREDENTIALS_UNSAFE]: The documentation suggests storing sensitive wallet credentials in a way that is susceptible to local exposure.
Evidence: Instructs users to create a plaintext password file at ~/.foundry/.password to store keystore passwords.
Mitigation: The skill provides warnings that this is for development only and implements a trap in its script to unset private key environment variables upon exit.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources and interpolates user input into shell commands, creating a potential attack surface.
Ingestion points: Fetches swap routes and transaction data from KyberSwap Aggregator and Token APIs.
Capability inventory: High-privilege actions including network access, file reads, and financial transaction broadcasting via cast.
Sanitization: Implements strict regex validation for all script parameters (addresses, amounts, symbols, chain slugs) to prevent command injection. It also uses a redaction mechanism to remove hex strings and private keys from script error outputs before they are seen by the agent.
Boundary markers: Requires a specific affirmative response ("yes", "proceed") from the user after a final warning prompt before executing the dangerous transaction logic.

swap-execute-fast