The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell scripts and the Foundry 'cast' CLI to prepare and broadcast blockchain transactions. It strictly validates all user-provided inputs, such as amounts, addresses, and chain slugs, using regular expressions to ensure they conform to expected formats and to prevent command injection vulnerabilities.
[CREDENTIALS_UNSAFE]: The skill manages sensitive wallet information, including private keys and keystore passwords. It follows security best practices by unsetting the private key environment variable upon completion, enforcing restrictive file permissions (chmod 600) on local password files, and implementing a robust redaction mechanism that filters hex strings and private key arguments from script output and error logs.
[EXTERNAL_DOWNLOADS]: The skill's documentation and internal error messages provide instructions for installing the Foundry toolkit using a 'curl | bash' command from paradigm.xyz. Paradigm is a recognized and reputable technology firm in the blockchain space, and the skill does not automatically execute this external script.
[PROMPT_INJECTION]: The skill implements a verbal confirmation gate for its 'fast' execution path, ensuring that the user explicitly consents to the immediate broadcast of transactions. It also enforces a $1,000 USD safety threshold to limit the risk associated with unreviewed trades.
[PROMPT_INJECTION]: The skill interacts with external token and aggregator APIs. To mitigate indirect prompt injection or data poisoning risks, it verifies the transaction's destination router address against a known-good hardcoded value. Evidence Chain: Ingestion (KyberSwap Aggregator and Token APIs in fast-swap.sh); Boundary markers (Verbal confirmation required in SKILL.md); Capability (Blockchain transaction execution via cast send); Sanitization (Regex input validation and router address integrity check in execute-swap.sh).

swap-execute-fast